Papers by Yejin Choi
Guardrails and Security for LLMs: Safe, Secure and Controllable Steering of LLM Applications (2025.acl-tutorials)
Copied to clipboard
Traian Rebedea, Leon Derczynski, Shaona Ghosh, Makesh Narsimhan Sreedhar, Faeze Brahman, Liwei Jiang, Bo Li, Yulia Tsvetkov, Christopher Parisien, Yejin Choi
| Challenge: | Pretrained generative models provide novel ways for users to interact with computers. |
| Approach: | This tutorial provides an overview of key guardrail mechanisms developed for LLMs along with evaluation methodologies and a detailed security assessment protocol. |
| Outcome: | This tutorial provides an overview of key guardrail mechanisms developed for LLMs, along with evaluation methodologies and a detailed security assessment protocol. |
Contrastive Explanations for Model Interpretability (2021.emnlp-main)
Copied to clipboard
| Challenge: | Existing methods for producing model explanations seek all causal factors at once, making them difficult to comprehend. |
| Approach: | They propose a method to produce contrastive explanations in the latent space . they use attribution and token/span attribution to produce models that consider only contrastive reasoning . |
| Outcome: | The proposed method allows model behavior to consider only contrastive reasoning . it also uncovers which aspects of the input are useful for and against particular decisions . |
Cosmos QA: Machine Reading Comprehension with Contextual Commonsense Reasoning (D19-1)
Copied to clipboard
| Challenge: | Existing reading comprehension datasets focus on factual and literal understanding of context paragraphs, but our dataset focuses on reading between the lines over a diverse collection of everyday narratives. |
| Approach: | They propose a large-scale dataset that requires commonsense-based reading comprehension, formulated as multiple-choice questions. |
| Outcome: | The proposed architecture improves over the baselines of existing reading comprehension datasets and shows a significant gap between machine (68.4%) and human performance (94%). |
Thinking Like a Skeptic: Defeasible Inference in Natural Language (2020.findings-emnlp)
Copied to clipboard
Rachel Rudinger, Vered Shwartz, Jena D. Hwang, Chandra Bhagavatula, Maxwell Forbes, Ronan Le Bras, Noah A. Smith, Yejin Choi
| Challenge: | Defeasible inference is a mode of reasoning in which an inference may be weakened or overturned in light of new evidence. |
| Approach: | They propose a dataset for defeasible inference in natural language that includes extensions to existing inference datasets. |
| Outcome: | Defeasible NLI extends existing datasets for defeaasibility inference in natural language . generative models can weaken or strengthen inferences up to 68% of the time, it shows . |
proScript: Partially Ordered Scripts Generation (2021.findings-emnlp)
Copied to clipboard
| Challenge: | Scripts represent structured commonsense knowledge about prototypical events in everyday situations/scenarios such as bake a cake. |
| Approach: | They collect 6.4k crowdsourced partially ordered scripts and develop models that combine language generation and graph structure prediction to generate scripts. |
| Outcome: | The proposed models perform well on two tasks: edge prediction and script generation. |
Commonsense Knowledge Transfer for Pre-trained Language Models (2023.findings-acl)
Copied to clipboard
| Challenge: | Recent advances in pre-trained language models have transformed the landscape of natural language processing. |
| Approach: | They propose a framework to transfer commonsense knowledge stored in a neural commonsensing model to a general-purpose pre-trained language model. |
| Outcome: | Empirical results show that the proposed framework improves the model’s performance on downstream tasks that require commonsense reasoning. |
Modeling Naive Psychology of Characters in Simple Commonsense Stories (P18-1)
Copied to clipboard
| Challenge: | Understanding a narrative requires reasoning about the causal links between the events in the story and the mental states of the characters, even when those relationships are not explicitly stated. |
| Approach: | They propose a new annotation framework to explain naive psychology of story characters as fully-specified chains of mental states with respect to motivations and emotional reactions. |
| Outcome: | The proposed framework provides a baseline performance on several new tasks suggesting avenues for future research. |
A Call for Clarity in Beam Search: How It Works and When It Stops (2024.lrec-main)
Copied to clipboard
| Challenge: | Empirical results show that a modified beam decoding implementation improves decoding performance of strong, neural language generation models. |
| Approach: | They propose a modification to a beam decoding implementation that generalizes the stopping criterion and provides flexibility to the depth of search. |
| Outcome: | The proposed method improves decoding performance of strong models on news text summarization and machine translation over diverse language pairs with negligible inference slowdown. |
Discourse Understanding and Factual Consistency in Abstractive Summarization (2021.eacl-main)
Copied to clipboard
Saadia Gabriel, Antoine Bosselut, Jeff Da, Ari Holtzman, Jan Buys, Kyle Lo, Asli Celikyilmaz, Yejin Choi
| Challenge: | Existing abstractive summarization models often hallucinate information or generate factually incorrect summaries. |
| Approach: | They propose a general framework for abstractive summarization with factual consistency and distinct modeling of the narrative flow in an output summary. |
| Outcome: | The proposed framework generates abstracts with factual consistency and coherence significantly better than baselines. |
Twist Decoding: Diverse Generators Guide Each Other (2022.emnlp-main)
Copied to clipboard
Jungo Kasai, Keisuke Sakaguchi, Ronan Le Bras, Hao Peng, Ximing Lu, Dragomir Radev, Yejin Choi, Noah A. Smith
| Challenge: | Using a variety of language generation models, ensembling models is challenging during inference. |
| Approach: | They propose a method that decodes text models that do not assume a shared vocabulary, tokenization or generation order. |
| Outcome: | The proposed method outperforms models decoded in isolation over various scenarios. |
I Cast Detect Thoughts: Learning to Converse and Guide with Intents and Theory-of-Mind in Dungeons and Dragons (2023.acl-long)
Copied to clipboard
Pei Zhou, Andrew Zhu, Jennifer Hu, Jay Pujara, Xiang Ren, Chris Callison-Burch, Yejin Choi, Prithviraj Ammanabrolu
| Challenge: | Existing dialogue agents, while able to produce human-like responses, often do not model goal-driven and grounded language interactions. |
| Approach: | They propose to decompose and model teacher-student natural language interactions into (1) the DM’s intent to guide players toward a given goal; (2) the dm’s guidance utterance to the players expressing this intent; (3) a theory-of-mind model that anticipates the players’ reaction to the guidance one turn into the future. |
| Outcome: | The proposed task is based on a goal-driven and grounded environment with a teacher-student interaction model and theory-of-mind model. |
Inference-Time Policy Adapters (IPA): Tailoring Extreme-Scale LMs without Fine-tuning (2023.emnlp-main)
Copied to clipboard
Ximing Lu, Faeze Brahman, Peter West, Jaehun Jung, Khyathi Chandu, Abhilasha Ravichander, Prithviraj Ammanabrolu, Liwei Jiang, Sahana Ramnath, Nouha Dziri, Jillian Fisher, Bill Lin, Skyler Hallinan, Lianhui Qin, Xiang Ren, Sean Welleck, Yejin Choi
| Challenge: | Extreme-scale language models have shown exceptional performance on a variety of language tasks, but the degree of control offered by these models through pure prompting is limited. |
| Approach: | They propose an inference-time policy adapter which tailors a large base model without fine-tuning it. |
| Outcome: | The proposed model outperforms baseline methods on five challenging text generation tasks and even over GPT-4. |
NeuroLogic Decoding: (Un)supervised Neural Text Generation with Predicate Logic Constraints (2021.naacl-main)
Copied to clipboard
| Challenge: | Conditional text generation often requires lexical constraints, i.e., which words should or shouldn't be included in the output text. |
| Approach: | They propose an algorithm that enables neural language models to generate fluent text while satisfying complex lexical constraints. |
| Outcome: | The proposed algorithm outperforms existing methods on four benchmarks and shows that it handles any set of lexical constraints expressible under predicate logic while its asymptotic runtime is equivalent to conventional beam search. |
MathQA: Towards Interpretable Math Word Problem Solving with Operation-Based Formalisms (N19-1)
Copied to clipboard
| Challenge: | Existing datasets in this domain do not offer precise operational annotations over diverse problem types due to noise and lack of formal operation-based representations. |
| Approach: | They propose a representation language to map problems to their operation programs . they also introduce an interpretable neural math problem solver . |
| Outcome: | The proposed model outperforms baseline models and the AQUA-RAT dataset on the AQuA-rat dataset. |
Unsupervised Commonsense Question Answering with Self-Talk (2020.emnlp-main)
Copied to clipboard
| Challenge: | Current systems rely on pre-trained language models or external knowledge bases to incorporate additional relevant knowledge. |
| Approach: | They propose an unsupervised framework based on self-talk to improve commonsense performance by asking language models to ask information seeking questions. |
| Outcome: | Empirical results show that the proposed framework improves on four out of six commonsense benchmarks and competes with models that obtain knowledge from external KBs. |
Impossible Distillation for Paraphrasing and Summarization: How to Make High-quality Lemonade out of Small, Low-quality Model (2024.naacl-long)
Copied to clipboard
Jaehun Jung, Peter West, Liwei Jiang, Faeze Brahman, Ximing Lu, Jillian Fisher, Taylor Sorensen, Yejin Choi
| Challenge: | Impossible Distillation is a framework for paraphrasing and sentence summarization that can be trained from a low-quality teacher model. |
| Approach: | They propose a framework that distills a high-quality dataset from a low-quality teacher . they hypothesize and verify the paraphrastic proximity intrinsic to pre-trained LMs . |
| Outcome: | The proposed framework outperforms baseline models on unconstrained paraphrase generation and sentence summarization benchmarks. |
Ultra-Fine Entity Typing (P18-1)
Copied to clipboard
| Challenge: | Experimental results show that a model that can predict ultra-fine types can be crowd-sourced . head words indicate the type of the noun phrases they appear in, and are important for context-sensitive tasks . |
| Approach: | They propose a task where sentences are given with an entity mention . they introduce a new type of distant supervision: head words, which indicate the type of noun phrases they appear in. |
| Outcome: | The proposed model can predict ultra-fine types at varying granularity and performs well on a fine-grained entity typing benchmark. |
FANToM: A Benchmark for Stress-testing Machine Theory of Mind in Interactions (2023.emnlp-main)
Copied to clipboard
| Challenge: | Existing evaluations for theory of mind (ToM) use passive narratives that lack interactivity. |
| Approach: | They propose a benchmark to stress-test ToM within information-asymmetric conversational contexts via question answering. |
| Outcome: | The proposed benchmark is challenging for state-of-the-art language models, which perform significantly worse than humans even with chain-of thought reasoning or fine-tuning. |
Generative Data Augmentation for Commonsense Reasoning (2020.findings-emnlp)
Copied to clipboard
Yiben Yang, Chaitanya Malaviya, Jared Fernandez, Swabha Swayamdipta, Ronan Le Bras, Ji-Ping Wang, Chandra Bhagavatula, Yejin Choi, Doug Downey
| Challenge: | Recent advances in commonsense reasoning depend on large-scale human-authored training data. |
| Approach: | They propose a generative data augmentation technique that augments human-authored training data by using pretrained language models. |
| Outcome: | The proposed technique outperforms existing methods on commonsense reasoning benchmarks and enhances out-of-distribution generalization. |
“I’m Not Mad”: Commonsense Implications of Negation and Contradiction (2021.naacl-main)
Copied to clipboard
| Challenge: | a new commonsense knowledge graph for negated and contradicted events is developed to help humans reason about their underlying causes and effects. |
| Approach: | They propose a new commonsense knowledge graph with 624K if-then rules focusing on negated and contradictory events. |
| Outcome: | The proposed model can be used to analyze negated and contradicted statements in natural language. |
Counterfactual Story Reasoning and Generation (D19-1)
Copied to clipboard
| Challenge: | a desired property of AI systems is counterfactual reasoning: ability to predict causal changes in future events. |
| Approach: | They propose to rewrite a short story and a counterfactual event to make it compatible with the given counterfact. |
| Outcome: | The proposed task requires deep understanding of causal narrative chains and counterfactual invariance . the proposed dataset includes 81,407 counterfact "branches" without a rewritten storyline . |
WANLI: Worker and AI Collaboration for Natural Language Inference Dataset Creation (2022.findings-emnlp)
Copied to clipboard
| Challenge: | Existing datasets are often flooded with repetitive and spurious patterns, leading to a lack of linguistic diversity. |
| Approach: | They propose a method that uses cartography to automatically identify and filter examples that demonstrate challenging reasoning patterns and then automatically compose new ones with similar patterns. |
| Outcome: | The proposed approach improves performance on eight out-of-domain test sets compared to training on the 4x larger MultiNLI dataset. |
It’s not Rocket Science: Interpreting Figurative Language in Narratives (2022.tacl-1)
Copied to clipboard
| Challenge: | Existing text representations by design rely on compositionality, while figurative language is often non-compositional. |
| Approach: | They propose to use a pre-trained language model to interpret figurative language types to adopt human strategies for interpreting figurativ language types: inferring meaning from context and relying on constituent words’ literal meanings. |
| Outcome: | The proposed models perform significantly worse than humans on discriminative and generative tasks, bridging the gap from human performance. |
DExperts: Decoding-Time Controlled Text Generation with Experts and Anti-Experts (2021.acl-long)
Copied to clipboard
Alisa Liu, Maarten Sap, Ximing Lu, Swabha Swayamdipta, Chandra Bhagavatula, Noah A. Smith, Yejin Choi
| Challenge: | Decoding-time Experts is a decoding- time method for controlled text generation . it combines a pretrained language model with "expert" LMs and/or "anti-expert" experts . |
| Approach: | They propose a decoding-time method that combines a pretrained language model with "expert" LMs and/or "anti-expert" experts to generate controlled text. |
| Outcome: | The proposed method outperforms existing controllable generation methods on automatic and human evaluations. |
What Makes it Ok to Set a Fire? Iterative Self-distillation of Contexts and Rationales for Disambiguating Defeasible Social and Moral Situations (2023.findings-emnlp)
Copied to clipboard
Kavel Rao, Liwei Jiang, Valentina Pyatkin, Yuling Gu, Niket Tandon, Nouha Dziri, Faeze Brahman, Yejin Choi
| Challenge: | Moral or ethical judgments rely heavily on the contexts in which they occur . a student model that produces defeasible contexts with improved validity, diversity, and defasibility is superior to intermediate student models . |
| Approach: | a new study uses a student model to provide contextualizations that make an action morally acceptable . the model is based on a dataset of 115K defeasible moral actions rated highly by human annotators . |
| Outcome: | The proposed model outperforms all intermediate models in a high-quality dataset . the model is based on 1.2M entries of contextualizations and rationales for 115K moral actions . |
Conversing by Reading: Contentful Neural Conversation with On-demand Machine Reading (P19-1)
Copied to clipboard
Lianhui Qin, Michel Galley, Chris Brockett, Xiaodong Liu, Xiang Gao, Bill Dolan, Yejin Choi, Jianfeng Gao
| Challenge: | a new approach to contentful neural conversation is proposed . end-to-end models are effective in learning fluent responses, but their responses are often vacuous and uninformative. |
| Approach: | They propose a model that provides the conversation model with relevant text on the fly as a source of external knowledge. |
| Outcome: | The proposed model improves the informativeness and diversity of generated output compared to previous methods. |
CommonGen: A Constrained Text Generation Challenge for Generative Commonsense Reasoning (2020.findings-emnlp)
Copied to clipboard
| Challenge: | Recent studies show that pre-trained language models perform well on commonsense-reasoning benchmark datasets, but building machines with commonsence to compose plausible sentences remains challenging. |
| Approach: | They propose a constrained text generation task for generative commonsense reasoning that generates a coherent sentence using common concepts. |
| Outcome: | The proposed task generates a coherent sentence describing an everyday scenario using common concepts over 35k concept-sets. |
“You Are An Expert Linguistic Annotator”: Limits of LLMs as Analyzers of Abstract Meaning Representation (2023.findings-emnlp)
Copied to clipboard
| Challenge: | Large language models (LLMs) demonstrate proficiency and fluency in the use of language, but do they have the linguistic knowledge to serve as an expert linguistic annotator? |
| Approach: | They examine the successes and limitations of large language models using the Abstract Meaning Representation (AMR) parsing formalism. |
| Outcome: | The proposed models can reproduce the basic format of AMR, as well as some core event, argument, and modifier structure, but they have virtually no fully accurate parses. |
Socratic-MCTS: Test-Time Visual Reasoning by Asking the Right Questions (2025.emnlp-main)
Copied to clipboard
| Challenge: | Recent research in vision-language models has centered around the possibility of equipping them with implicit long-form chain-of-thought reasoning via distillation and reinforcement learning. |
| Approach: | They propose a Monte Carlo Tree Search-inspired algorithm that injects subquestion–subanswer pairs into the model’s output stream to elicit hidden knowledge and induce long reasoning traces. |
| Outcome: | The proposed method yields a 2% improvement on MMMU-PRO, including a significant 9% gain in Liberal Arts. |
Bidimensional Leaderboards: Generate and Evaluate Language Hand in Hand (2022.naacl-main)
Copied to clipboard
Jungo Kasai, Keisuke Sakaguchi, Ronan Le Bras, Lavinia Dunagan, Jacob Morrison, Alexander Fabbri, Yejin Choi, Noah A. Smith
| Challenge: | Recent advances on models and metrics should benefit and inform each other, authors argue . bidimensional leaderboards allow for fast, accurate evaluation of language generation models . |
| Approach: | They propose a bidimensional leaderboard that tracks progress in language generation models and metrics for their evaluation. |
| Outcome: | The proposed leaderboards track progress in language generation models and metrics for their evaluation. |
Challenges in Automated Debiasing for Toxic Language Detection (2021.eacl-main)
Copied to clipboard
| Challenge: | Existing methods for debiasing toxic language data are limited in their ability to prevent biased behavior in toxic language detection systems. |
| Approach: | They propose to debiase toxic language detection models using lexical and dialectal markers using synthetic labels instead of traditional methods. |
| Outcome: | The proposed method reduces dialectal associations with toxicity despite the use of synthetic labels . |
NeuroComparatives: Neuro-Symbolic Distillation of Comparative Knowledge (2024.findings-naacl)
Copied to clipboard
| Challenge: | Comparative knowledge is an essential component of our world knowledge, yet understudied in prior literature. |
| Approach: | They propose a framework for comparative knowledge distillation overgenerated from language models . they use a corpus of 8.8M comparisons over 1.74M entity pairs to acquire comparative information . |
| Outcome: | The proposed framework acquires comparative knowledge between everyday objects . human evaluations show that it outperforms existing resources in terms of validity . |
OLMoTrace: Tracing Language Model Outputs Back to Trillions of Training Tokens (2025.acl-demo)
Copied to clipboard
Jiacheng Liu, Taylor Blanton, Yanai Elazar, Sewon Min, Yen-Sung Chen, Arnavi Chheda-Kothary, Huy Tran, Byron Bischoff, Eric Marsh, Michael Schmitz, Cassidy Trier, Aaron Sarnat, Jenna James, Jon Borchardt, Bailey Kuehl, Evie Yu-Yen Cheng, Karen Farley, Taira Anderson, David Albright, Carissa Schoenick, Luca Soldaini, Dirk Groeneveld, Rock Yuren Pang, Pang Wei Koh, Noah A. Smith, Sophie Lebrecht, Yejin Choi, Hannaneh Hajishirzi, Ali Farhadi, Jesse Dodge
| Challenge: | tracing language models' outputs back to training data is a problem because they are trained on text corpora with trillions of tokens . existing methods for tracers have not been scaled to work within this multi-trillion-token setting . |
| Approach: | They propose a system that traces language models' outputs verbatim back to training data . OLMOTRACE retrieves documents from the model's training data that contain exact matches . |
| Outcome: | The proposed system can find verbatim matches between LM output and training data . it can be used to explore fact checking, hallucination, and creativity of language models . |
Selective “Selective Prediction”: Reducing Unnecessary Abstention in Vision-Language Reasoning (2024.findings-acl)
Copied to clipboard
Tejas Srinivasan, Jack Hessel, Tanmay Gupta, Bill Yuchen Lin, Yejin Choi, Jesse Thomason, Khyathi Chandu
| Challenge: | ReCoVERR reduces the over-abstention of a vision-language system with low tolerance for inaccurate predictions without increasing the error rate of the system’s predictions. |
| Approach: | They propose an inference-time algorithm to reduce the over-abstention of a selective vision-language system without increasing the error rate of the system’s predictions. |
| Outcome: | ReCoVERR reduces the over-abstention of a vision-language system without increasing the error rate of the system’s predictions. |
Reflective Decoding: Beyond Unidirectional Generation with Off-the-Shelf Language Models (2021.acl-long)
Copied to clipboard
| Challenge: | Existing methods for generating text are unsupervised and require supervision. |
| Approach: | They propose an unsupervised method that uses two off-the-shelf pretrained LMs in opposite directions to apply them to non-sequential tasks. |
| Outcome: | The proposed method outperforms strong unsupervised baselines on paraphrasing and abductive text infilling. |
BottleSum: Unsupervised and Self-supervised Sentence Summarization using the Information Bottleneck Principle (D19-1)
Copied to clipboard
| Challenge: | Existing approaches to extractive and abstractive summarization rely on large-scale parallel corpora of input text and output summaries for direct supervision. |
| Approach: | They propose an unsupervised approach to sentence summarization using the Information Bottleneck principle. |
| Outcome: | The proposed method outperforms unsupervised models on automatic metrics and human evaluation along multiple attributes. |
How to Train Your Fact Verifier: Knowledge Transfer with Multimodal Open Models (2024.findings-emnlp)
Copied to clipboard
Jaeyoung Lee, Ximing Lu, Jack Hessel, Faeze Brahman, Youngjae Yu, Yonatan Bisk, Yejin Choi, Saadia Gabriel
| Challenge: | a growing influx of misinformation across news and social media is hampered by outdated foundation model training data. |
| Approach: | They propose to use large language models to scale up online policing mechanisms . they evaluate foundation model performance without continual updating . |
| Outcome: | The proposed model can improve performance without continual updating . the proposed model improves on two widely used benchmarks . |
SWAG: A Large-Scale Adversarial Dataset for Grounded Commonsense Inference (D18-1)
Copied to clipboard
| Challenge: | a new dataset presents a task of grounded commonsense inference, unifying natural language inference and commonsensical reasoning. |
| Approach: | They propose a procedure that constructs a de-biased dataset by iteratively training stylistic classifiers and using them to filter the data. |
| Outcome: | The proposed procedure oversamples a de-biased dataset using state-of-the-art language models . human models struggle on the proposed procedure, indicating significant opportunities for future research. |
BotPercent: Estimating Bot Populations in Twitter Communities (2023.findings-emnlp)
Copied to clipboard
| Challenge: | Existing approaches to bot detection are agnostic to social environments the bots operate in . however, standard approaches are not a good fit for the social environments they operate in. |
| Approach: | They propose a method that estimates the percentage of Twitter bots given a community . they use Twitter bot detection datasets and feature-, text-, and graph-based models adjusted to a particular community based on Twitter . |
| Outcome: | The proposed method achieves state-of-the-art in community-level Twitter bot detection across balanced and imbalanced class distribution settings. |
HellaSwag: Can a Machine Really Finish Your Sentence? (P19-1)
Copied to clipboard
| Challenge: | Existing commonsense models struggle to perform inferences that are trivial for humans, but are often misclassified by state-of-the-art models. |
| Approach: | They propose a dataset that is adversarial to state-of-the-art commonsense reasoning and use it to build a model that is surprisingly robust. |
| Outcome: | The proposed dataset is compared with existing models and scaled up towards a critical 'Goldilocks zone' wherein generated text is ridiculous to humans, yet often misclassified by state-of-the-art models. |
Biased LLMs can Influence Political Decision-Making (2025.acl-long)
Copied to clipboard
Jillian Fisher, Shangbin Feng, Robert Aron, Thomas Richardson, Yejin Choi, Daniel W Fisher, Jennifer Pan, Yulia Tsvetkov, Katharina Reinecke
| Challenge: | Recent studies have found that biased LLMs can influence decisions in areas such as medical classifications and educational hiring. |
| Approach: | They conducted two interactive experiments on partisan bias in large language models while completing tasks with either a biased liberal, biased conservative, or unbiased control model. |
| Outcome: | The results show that prior knowledge of AI is weakly correlated with a reduction of the bias, suggesting that AI education can be crucial for mitigating bias effects. |
Are Machine Rationales (Not) Useful to Humans? Measuring and Improving Human Utility of Free-text Rationales (2023.acl-long)
Copied to clipboard
Brihi Joshi, Ziyi Liu, Sahana Ramnath, Aaron Chan, Zhewei Tong, Shaoliang Nie, Qifan Wang, Yejin Choi, Xiang Ren
| Challenge: | Existing metrics like task performance of the LM generating the rationales or similarity between generated and gold rationale are not good indicators of their human utility. |
| Approach: | They propose to use a large language model to generate rationales with better human utility by estimating its conciseness and novelty. |
| Outcome: | The proposed model can measure human utility to a better extent by estimating its usefulness in answering similar unseen instances. |
Connecting the Dots between Audio and Text without Parallel Data through Visual Knowledge Transfer (2022.naacl-main)
Copied to clipboard
| Challenge: | Existing methods for learning audio-text connections rely on parallel audio- text data . a new approach allows for the representation of environmental soundscapes without using parallel data - a challenge for many applications . |
| Approach: | They propose a model that induces Audio-Text alignment without using parallel audio-text data. |
| Outcome: | The proposed model outperforms the current state-of-the-art for audio classification tasks with no audio-text data by 2.2% on the ESC50 and US8K tasks. |
NovaCOMET: Open Commonsense Foundation Models with Symbolic Knowledge Distillation (2023.findings-emnlp)
Copied to clipboard
Peter West, Ronan Bras, Taylor Sorensen, Bill Lin, Liwei Jiang, Ximing Lu, Khyathi Chandu, Jack Hessel, Ashutosh Baheti, Chandra Bhagavatula, Yejin Choi
| Challenge: | a new commonsense knowledge model, NovaCOMET, combines knowledge and general task models. |
| Approach: | They propose an open commonsense knowledge model that combines knowledge and general task models. |
| Outcome: | The proposed model matches or exceeds existing knowledge models on commonsense reasoning tasks. |
Aligning to Social Norms and Values in Interactive Narratives (2022.naacl-main)
Copied to clipboard
| Challenge: | Social value alignment is the ability to create agents that act in alignment with socially beneficial norms and values in interactive narratives or text-based games. |
| Approach: | They introduce a game-value ALignment agent that uses social commonsense to restrict its action space to actions that are aligned with socially beneficial values. |
| Outcome: | The proposed agent improves state-of-the-art task performance by 4% while reducing the frequency of socially harmful behaviors by 25% compared to strong contemporary value alignment approaches. |
CulturalBench: A Robust, Diverse and Challenging Benchmark for Measuring LMs’ Cultural Knowledge Through Human-AI Red-Teaming (2025.acl-long)
Copied to clipboard
Yu Ying Chiu, Liwei Jiang, Bill Yuchen Lin, Chan Young Park, Shuyue Stella Li, Sahithya Ravi, Mehar Bhatia, Maria Antoniak, Yulia Tsvetkov, Vered Shwartz, Yejin Choi
| Challenge: | CulturalBench is a set of 1,696 human-written and human-verified questions to assess LMs’ cultural knowledge covering 45 global regions including underrepresented ones like Bangladesh, Zimbabwe, and Peru. |
| Approach: | They construct a set of 1,696 human-written and human-verified questions to assess LMs' cultural knowledge, covering 45 global regions including underrepresented ones like Bangladesh, Zimbabwe, and Peru. |
| Outcome: | The proposed model outperforms other models across cultures, while underperforming on questions related to North Africa, South America and Middle East. |
The Risk of Racial Bias in Hate Speech Detection (P19-1)
Copied to clipboard
| Challenge: | Annotators’ insensitivity to differences in dialect can lead to racial bias in automatic hate speech detection models, potentially amplifying harm against minority populations. |
| Approach: | They propose *dialect* and *race priming* as ways to reduce the racial bias in hate speech detection models by detecting differences in dialects in annotated tweets. |
| Outcome: | The proposed models acquire and propagate these biases, such that AAE tweets and tweets by self-identified African Americans are up to two times more likely to be labelled as offensive compared to others. |
RealToxicityPrompts: Evaluating Neural Toxic Degeneration in Language Models (2020.findings-emnlp)
Copied to clipboard
| Challenge: | Pretrained neural language models (LMs) are prone to generating toxic language . we investigate the extent to which pretrained LMs can be prompted to generate toxic language. |
| Approach: | They propose to use a dataset to test the extent to which pretrained LMs can generate toxic language . they use toxicity-based text generation algorithms to generate toxic text from natural prompts . |
| Outcome: | The proposed model can degenerate into toxic text even from seemingly innocuous prompts. |
Misinfo Reaction Frames: Reasoning about Readers’ Reactions to News Headlines (2022.acl-long)
Copied to clipboard
Saadia Gabriel, Skyler Hallinan, Maarten Sap, Pemi Nguyen, Franziska Roesner, Eunsol Choi, Yejin Choi
| Challenge: | Empirical results confirm that it is indeed possible for neural models to predict the prominent patterns of readers’ reactions to previously unseen news headlines. |
| Approach: | They propose a pragmatic formalism for modeling how readers might react to a news headline . they propose 'misinfo' frames, which can be used to model reader perceptions of news reliability . |
| Outcome: | The proposed model can predict readers' reactions to previously unseen headlines. |
NeuroCounterfactuals: Beyond Minimal-Edit Counterfactuals for Richer Data Augmentation (2022.findings-emnlp)
Copied to clipboard
| Challenge: | Existing approaches to produce counterfactuals rely on small perturbations via minimal edits, resulting in simplistic changes. |
| Approach: | They propose a novel approach to produce counterfactuals that allow for larger edits and linguistic diversity while still bearing similarity to the original document. |
| Outcome: | The proposed approach outperforms existing methods for generalizing natural language models under select settings. |
ALPACA AGAINST VICUNA: Using LLMs to Uncover Memorization of LLMs (2025.naacl-long)
Copied to clipboard
Aly M. Kassem, Omar Mahmoud, Niloofar Mireshghallah, Hyunwoo Kim, Yulia Tsvetkov, Yejin Choi, Sherif Saad, Santu Rana
| Challenge: | Existing studies have shown that pre-trained LLMs emit training data up to 150 more often than in regular operation. |
| Approach: | They propose a black-box prompt optimization method where an attacker LLM agent uncovers higher levels of memorization in a victim agent . |
| Outcome: | The proposed method shows 23.7% more overlap with training data compared to state-of-the-art baselines. |
Do Androids Laugh at Electric Sheep? Humor “Understanding” Benchmarks from The New Yorker Caption Contest (2023.acl-long)
Copied to clipboard
Jack Hessel, Ana Marasovic, Jena D. Hwang, Lillian Lee, Jeff Da, Rowan Zellers, Robert Mankoff, Yejin Choi
| Challenge: | Large neural networks can generate jokes, but do they really “understand” humor? a new challenge challenges AI models to match a joke to a cartoon, identify a winning caption, and explain why a winner is funny. |
| Approach: | They propose three tasks based on the New Yorker Cartoon Caption Contest . they aim to match a joke to a cartoon, identify a winning caption and explain why it's funny . |
| Outcome: | The proposed tasks are based on the New Yorker Cartoon Caption Contest . they include matching a joke to a cartoon, identifying a winning caption, and explaining why a funny caption is funny. |
Nemotron-CrossThink: Scaling Self-Learning beyond Math Reasoning (2026.eacl-long)
Copied to clipboard
Syeda Nahida Akter, Shrimai Prabhumoye, Matvei Novikov, Seungju Han, Ying Lin, Evelina Bakhturina, Eric Nyberg, Yejin Choi, Mostofa Patwary, Mohammad Shoeybi, Bryan Catanzaro
| Challenge: | Prior work has successfully applied Reinforcement Learning (RL) to mathematical reasoning, but generalization to broader domains remains challenging due to limited data and lack of verifiable rewards for unstructured domains. |
| Approach: | They propose a framework that integrates multi-domain corpora into RL training to improve generalization across diverse reasoning tasks. |
| Outcome: | The proposed framework improves generalization across diverse reasoning tasks. |
Infini-gram mini: Exact n-gram Search at the Internet Scale with FM-Index (2025.emnlp-main)
Copied to clipboard
| Challenge: | Modern language models are trained on text data downsampled from massive text corpora like Common Crawl. |
| Approach: | They propose an efficient and scalable system that can make petabyte-level text corpora searchable by using the FM-index data structure. |
| Outcome: | The proposed system indexes 83TB of Internet text in 99 days with a single 128-core CPU node (or 19 hours if using 137 such nodes). |
COMET: Commonsense Transformers for Automatic Knowledge Graph Construction (P19-1)
Copied to clipboard
| Challenge: | Existing studies on commonsense knowledge base construction only store loosely structured open-text descriptions of knowledge. |
| Approach: | They propose a commonsense knowledge base construction model that generates rich commonsensense descriptions in natural language. |
| Outcome: | The proposed models can generate rich and diverse commonsense descriptions in natural language. |
Recollection versus Imagination: Exploring Human Memory and Cognition via Neural Language Models (2020.acl-main)
Copied to clipboard
| Challenge: | a dataset of imagined and recalled stories is used to study the cognitive processes involved in storytelling, contrasting imagination and recollection of events. |
| Approach: | They use a dataset of 7,000 stories to study the cognitive processes involved in storytelling, contrasting imagination and recollection of events. |
| Outcome: | The proposed measures show that imagined stories have a substantially more linear narrative flow compared to recalled stories in which adjacent sentences are more disconnected. |
Agent Lumos: Unified and Modular Training for Open-Source Language Agents (2024.acl-long)
Copied to clipboard
Da Yin, Faeze Brahman, Abhilasha Ravichander, Khyathi Chandu, Kai-Wei Chang, Yejin Choi, Bill Yuchen Lin
| Challenge: | Lumos is a framework for training open-source agents on complex interactive tasks. |
| Approach: | They propose a framework for training open-source LLM-based agents called Lumos . Lumos features a learnable, unified and modular architecture with a planning module that learns high-level subgoal generation and a grounding module trained to translate these into the actions using various tools in the execution module. |
| Outcome: | The framework outperforms open-source agents on QA and web tasks. |
WildVis: Open Source Visualizer for Million-Scale Chat Logs in the Wild (2024.emnlp-demo)
Copied to clipboard
| Challenge: | Currently, the volume and complexity of chat logs makes it difficult to analyze individual conversations. |
| Approach: | They propose a tool that enables fast, versatile, and large-scale conversation analysis by combining search and visualization capabilities with a list of criteria. |
| Outcome: | The proposed tool can be extended to handle millions of chat logs and other datasets. |
Social IQa: Commonsense Reasoning about Social Interactions (D19-1)
Copied to clipboard
| Challenge: | Social IQa contains 38,000 multiple choice questions for probing emotional and social intelligence in a variety of everyday situations. |
| Approach: | They propose a crowdsourcing framework that collects commonsense questions along with correct and incorrect answers about social interactions. |
| Outcome: | The proposed framework mitigates stylistic artifacts in incorrect answers by asking workers to provide the right answer to a different but related question. |
Social Chemistry 101: Learning to Reason about Social and Moral Norms (2020.emnlp-main)
Copied to clipboard
| Challenge: | SOCIAL CHEMISTRY is a conceptual formalism to study people’s everyday social norms and moral judgments over a rich spectrum of real life situations described in natural language. |
| Approach: | They propose a new conceptual formalism to study people's everyday social norms and moral judgments over a rich spectrum of real life situations described in natural language. |
| Outcome: | The proposed model can be used to model people's everyday social norms and moral judgments over a rich spectrum of real life situations. |
Prompt Waywardness: The Curious Case of Discretized Interpretation of Continuous Prompts (2022.naacl-main)
Copied to clipboard
Daniel Khashabi, Xinxi Lyu, Sewon Min, Lianhui Qin, Kyle Richardson, Sean Welleck, Hannaneh Hajishirzi, Tushar Khot, Ashish Sabharwal, Sameer Singh, Yejin Choi
| Challenge: | Recent work shows the surprising power of continuous prompts to language models for controlled generation and solving a wide range of tasks. |
| Approach: | They propose to extract a discrete (textual) interpretation of continuous prompts faithful to the problem they solve. |
| Outcome: | The proposed model can find prompts that solve a task while being projected to an arbitrary text with a smaller drop in accuracy. |
TuringAdvice: A Generative and Dynamic Evaluation of Language Use (2021.naacl-main)
Copied to clipboard
| Challenge: | Empirical results show that today’s language models struggle at TuringAdvice . language models are getting ever-larger, and are being trained on ever-increasing quantities of text . |
| Approach: | They propose a task task that requires models to generate helpful advice in natural language. |
| Outcome: | The proposed model outperforms even multibillion parameter models on 600k in-domain training examples. |
Do MLLMs Capture How Interfaces Guide User Behavior? A Benchmark for Multimodal UI/UX Design Understanding (2026.acl-long)
Copied to clipboard
Jaehyun Jeon, Min Soo Kim, Janghan Yoon, Sumin Shim, Yejin Choi, Hanbin Kim, Dae Hyun Kim, Youngjae Yu
| Challenge: | Recent studies focus on surface-level features, overlooking how design choices influence user behavior at scale. |
| Approach: | They propose a benchmark for multimodal understanding of how UI/UX design affects user behavior built on 300 real-world UI image pairs from industry A/B tests. |
| Outcome: | The proposed benchmarks show that models exhibit limited understanding of the behavioral impact of UI/UX design. |
Conversational Multi-Hop Reasoning with Neural Commonsense Knowledge and Symbolic Logic Rules (2021.emnlp-main)
Copied to clipboard
| Challenge: | Currently, conversational agents lack commonsense reasoning, preventing them from engaging in rich conversations with humans. |
| Approach: | They propose a commonsense reasoning system that uncovers unstated presumptions from user commands satisfying a general template of if-(state), then-(action), because-(goal) They propose to use a transformer-based generative commons sense knowledge base as its source of background knowledge to extract multi-hop reasoning chains from the neural KB. |
| Outcome: | The proposed model achieves a 35% higher success rate than existing methods with human users. |
Clever Hans or Neural Theory of Mind? Stress Testing Social Reasoning in Large Language Models (2024.eacl-long)
Copied to clipboard
Natalie Shapira, Mosh Levy, Seyed Hossein Alavi, Xuhui Zhou, Yejin Choi, Yoav Goldberg, Maarten Sap, Vered Shwartz
| Challenge: | Recent work suggests that Large Language Models (LLMs) exhibit Neural Theory-of-Mind (N-ToM) however, prior work reached conflicting conclusions regarding those abilities. |
| Approach: | They examine the extent of Large Language Models’ N-ToM abilities through an extensive evaluation of 6 tasks and find that LLMs struggle with adversarial examples . |
| Outcome: | The proposed metrics show that LLMs exhibit certain N-ToM abilities, but this behavior is far from robust. |
Penguins Don’t Fly: Reasoning about Generics through Instantiations and Exceptions (2023.eacl-main)
Copied to clipboard
| Challenge: | Generics express generalizations about the world that are not universally true . commonsense knowledge bases encode some generic knowledge but rarely enumerate exceptions . |
| Approach: | They propose a framework informed by linguistic theory to generate exemplars for generics . they generate 19k exemplar cases for 650 generics and show they outperform a strong baseline . |
| Outcome: | The proposed framework outperforms a baseline framework by 12.8 precision points. |
Detoxifying Text with MaRCo: Controllable Revision with Experts and Anti-Experts (2023.acl-short)
Copied to clipboard
| Challenge: | Text detoxification can mitigate the harms of toxicity by rephrasing text to remove offensive meaning, but subtle toxicity remains challenging to tackle. |
| Approach: | They propose a text detoxification algorithm that combines controllable generation and text rewriting methods using a Product of Experts and autoencoder language models to find candidate words to mask and potentially replace. |
| Outcome: | The proposed method outperforms baselines on automatic metrics and is preferred 2.1 times more in human evaluation. |
Probing Factually Grounded Content Transfer with Factual Ablation (2022.findings-acl)
Copied to clipboard
| Challenge: | Despite recent success, large neural models often generate factually incorrect text . lack of a standard evaluation for factuality complicates factual grounded generation . |
| Approach: | They propose a method to measure factual consistency by presenting two evaluation sets . large pretrained models have shown impressive effectiveness at longstanding tasks . |
| Outcome: | The proposed method improves over strong baselines by presenting two evaluation sets. |
I2D2: Inductive Knowledge Distillation with NeuroLogic and Self-Imitation (2023.acl-long)
Copied to clipboard
Chandra Bhagavatula, Jena D. Hwang, Doug Downey, Ronan Le Bras, Ximing Lu, Lianhui Qin, Keisuke Sakaguchi, Swabha Swayamdipta, Peter West, Yejin Choi
| Challenge: | Empirical results suggest that scale is not the only way to build commonsense capabilities. |
| Approach: | They propose a commonsense distillation framework that can achieve a competitive level of commonsensing without relying on the benefits of scale. |
| Outcome: | The proposed framework breaks the dependence on the extreme-scale teacher model with two innovations: (1) the novel adaptation of NeuroLogic Decoding to enhance the generation quality of the weak, off-the-shelf language models, and (2) self-imitation learning to iteratively learn from the model’s own enhanced commonsense acquisition capabilities. |
Natural Language Rationales with Full-Stack Visual Reasoning: From Pixels to Semantic Frames to Commonsense Graphs (2020.findings-emnlp)
Copied to clipboard
| Challenge: | Existing models that use natural language rationales provide intuitive, higher-level explanations that are easily understandable by humans. |
| Approach: | They propose a model that generates free-text rationales by combining pretrained language models with object recognition, grounded visual semantic frames, and visual commonsense graphs. |
| Outcome: | The proposed model generates free-text rationales by combining pretrained language models with object recognition, grounded visual semantic frames, and visual commonsense graphs. |
PIGLeT: Language Grounding Through Neuro-Symbolic Interaction in a 3D World (2021.acl-long)
Copied to clipboard
Rowan Zellers, Ari Holtzman, Matthew Peters, Roozbeh Mottaghi, Aniruddha Kembhavi, Ali Farhadi, Yejin Choi
| Challenge: | PIGLeT model learns physical commonsense knowledge through interaction, then uses this knowledge to ground language. |
| Approach: | They propose a model that learns physical commonsense knowledge through interaction . they factorize PIGLeT into a physical dynamics model and a separate language model . |
| Outcome: | The proposed model outperforms a 100x larger, text-to-text approach in forecasting language . it can read a sentence, simulate neurally what might happen next, and communicate that result through a literal symbolic representation, or natural language. |
Transparent Human Evaluation for Image Captioning (2022.naacl-main)
Copied to clipboard
Jungo Kasai, Keisuke Sakaguchi, Lavinia Dunagan, Jacob Morrison, Ronan Le Bras, Yejin Choi, Noah A. Smith
| Challenge: | Recent work has demonstrated that image captioning is a complex task that requires a large amount of human input. |
| Approach: | They develop a human evaluation protocol for image captioning models based on machine- and human-generated captions on the MSCOCO dataset. |
| Outcome: | The proposed model improves CLIPScore, a recent metric that uses image features, and improves human judgments because it is more sensitive to recall. |
Neural Poetry Translation (N18-2)
Copied to clipboard
| Challenge: | despite recent advances in machine translation, automatic poetry translation remains a challenging problem. |
| Approach: | They propose a system that automatically translates a source text to an English poem . human evaluation of the translations ranks the quality as acceptable 78.2% of the time. |
| Outcome: | The proposed system always translates a source text to an English poem. human evaluation of the translations ranks the quality as acceptable 78.2% of the time. |
JAMDEC: Unsupervised Authorship Obfuscation using Constrained Decoding over Small Language Models (2024.naacl-long)
Copied to clipboard
| Challenge: | Existing methods to protect the identity and privacy of online authorship are lacking supervision data for diverse authorship and domains. |
| Approach: | They propose an unsupervised inference-time approach to authorship obfuscation that uses a user-controlled, inference time algorithm to oblige the authorship. |
| Outcome: | The proposed method outperforms state-of-the-art methods while performing competitively against a propriety model two orders of magnitudes larger. |
PowerTransformer: Unsupervised Controllable Revision for Biased Language Correction (2020.emnlp-main)
Copied to clipboard
| Challenge: | Unconscious biases continue to be prevalent in modern text and media, calling for algorithms that can assist writers with bias correction. |
| Approach: | They propose a new revision task that debiases text through the lens of connotation frames to correct implicit biases in character portrayals. |
| Outcome: | The proposed approach outperforms existing methods and ablations in the literature. |
Can Language Models Reason about Individualistic Human Values and Preferences? (2025.acl-long)
Copied to clipboard
| Challenge: | Existing methods and evaluation frameworks for achieving pluralistic alignment are limited by the diversity of people, which is pre-specified and coarsely categorized, papering over individuality. |
| Approach: | They propose to use a dataset transformed from the influential World Values Survey to study language models on the specific challenge of individualistic value reasoning. |
| Outcome: | The proposed model can predict individualistic values with accuracies between 55% and 65%, while a precise description of individualistic value judgments cannot be approximated only via demographic information. |
Modular Pluralism: Pluralistic Alignment via Multi-LLM Collaboration (2024.emnlp-main)
Copied to clipboard
Shangbin Feng, Taylor Sorensen, Yuhan Liu, Jillian Fisher, Chan Young Park, Yejin Choi, Yulia Tsvetkov
| Challenge: | Existing alignment paradigms for large language models learn an averaged human preference and struggle to model diverse preferences across cultures, demographics, and communities. |
| Approach: | They propose a modular framework that "plugs" into a base LLM a pool of smaller but specialized community LMs where models collaborate in distinct modes to support three modes of pluralism: Overton, steerable, and distributional. |
| Outcome: | The proposed framework “plugs into” a base LLM a pool of smaller but specialized community LMs, where models collaborate in distinct modes to support three modes of pluralism: Overton, steerable, and distributional. |
Generated Knowledge Prompting for Commonsense Reasoning (2022.acl-long)
Copied to clipboard
Jiacheng Liu, Alisa Liu, Ximing Lu, Sean Welleck, Peter West, Ronan Le Bras, Yejin Choi, Hannaneh Hajishirzi
| Challenge: | Existing methods for commonsense reasoning rely on high-quality knowledge, but they are often dominated by large-scale pretrained models that are fine-tuned on a target benchmark. |
| Approach: | They develop generated knowledge prompting which generates knowledge from a language model and provides it as additional input when answering a question. |
| Outcome: | The proposed method improves state-of-the-art models on four commonsense reasoning tasks. |
Discourse-Aware Neural Rewards for Coherent Text Generation (N18-1)
Copied to clipboard
| Challenge: | Existing approaches to train text generation models using cross-entropy loss do not always correlate well with achieving high scores on commonly used evaluation measures. |
| Approach: | They propose to use discourse-aware rewards to model cross-sentence ordering to approximate desired discourse structure to train a model of long, coherent text. |
| Outcome: | The proposed model produces more coherent and less repetitive text than models trained with cross-entropy or with commonly used scores as rewards. |
NaturalAdversaries: Can Naturalistic Adversaries Be as Effective as Artificial Adversaries? (2022.findings-emnlp)
Copied to clipboard
| Challenge: | Existing work has explored adversarial example generation for natural language understanding tasks, but these examples are often unrealistic and diverge from the real-world data distributions. |
| Approach: | They propose a framework for adversarial example generation that is effective at fooling a given classifier and a generative model based on the key tokens from the first stage. |
| Outcome: | The proposed framework generalizes across domains and offers insights for future research on improving robustness of neural text classification models. |
Symbolic Working Memory Enhances Language Models for Complex Rule Application (2024.emnlp-main)
Copied to clipboard
| Challenge: | Large Language Models (LLMs) excel in single-step rule application but struggle with multi-step deductive reasoning when rules are presented non-sequentially. |
| Approach: | They propose to augment LLMs with external working memory and introduce a neurosymbolic framework for rule application that stores facts and rules in both natural language and symbolic forms, enabling precise tracking. |
| Outcome: | The proposed framework iteratively performs symbolic rule grounding and LLM-based rule implementation. |
Commonsense Reasoning for Natural Language Processing (2020.acl-tutorials)
Copied to clipboard
| Challenge: | In this tutorial, we will outline the various types of commonsense knowledge and discuss techniques to gather and represent commonsence knowledge. |
| Approach: | This tutorial will provide researchers with the critical foundations and recent advances in commonsense representation and reasoning. |
| Outcome: | This tutorial will outline the various types of commonsense and discuss techniques to gather and represent commonsence knowledge while highlighting the challenges specific to this type of knowledge (e.g., reporting bias). |
CopyBench: Measuring Literal and Non-Literal Reproduction of Copyright-Protected Text in Language Model Generation (2024.emnlp-main)
Copied to clipboard
Tong Chen, Akari Asai, Niloofar Mireshghallah, Sewon Min, James Grimmelmann, Yejin Choi, Hannaneh Hajishirzi, Luke Zettlemoyer, Pang Wei Koh
| Challenge: | Existing studies focus on literal copying, but current methods reduce literal copy but not non-literal copying. |
| Approach: | They propose a benchmark to measure literal and non-literal copying in LMs . they use copyrighted fiction books as text sources to assess literal copying . |
| Outcome: | The proposed model measures literal and non-literal copying in copyrighted texts . large models show significantly more copying, with literal copying rates increasing . |
CLIPScore: A Reference-free Evaluation Metric for Image Captioning (2021.emnlp-main)
Copied to clipboard
| Challenge: | Image captioning relies on reference-based automatic evaluations, but references are expensive to collect and comparing against multiple human-authored captions is insufficient. |
| Approach: | They propose a reference-free metric that can be used for automatic caption evaluation without references. |
| Outcome: | The proposed model outperforms existing metrics on image-text compatibility and a reference-augmented version achieves even higher correlation with human judgements. |
Do Neural Language Models Overcome Reporting Bias? (2020.coling-main)
Copied to clipboard
| Challenge: | Recent studies show that pre-trained language models can overcome reporting bias by estimating the plausibility of rare but unspoken facts. |
| Approach: | They revisit the experiments conducted by Gordon and Van Durme (2013) . they find that pre-trained language models overestimate the very rare . |
| Outcome: | The proposed approach overestimates the rare at the expense of the rare, while minimizing reporting bias. |
SQuARe: A Large-Scale Dataset of Sensitive Questions and Acceptable Responses Created through Human-Machine Collaboration (2023.acl-long)
Copied to clipboard
Hwaran Lee, Seokhee Hong, Joonsuk Park, Takyoung Kim, Meeyoung Cha, Yejin Choi, Byoungpil Kim, Gunhee Kim, Eun-Ju Lee, Yong Lim, Alice Oh, Sangchul Park, Jung-Woo Ha
| Challenge: | Existing studies focus on coping with social harms that large language models pose . however, discussions on sensitive issues can become toxic even if the users are well-intentioned. |
| Approach: | They propose to use Korean dataset to test whether LLMs can generate offensive content and propagate prejudices. |
| Outcome: | The proposed dataset shows that acceptable response generation improves for HyperCLOVA and GPT-3. |
SODA: Million-scale Dialogue Distillation with Social Commonsense Contextualization (2023.emnlp-main)
Copied to clipboard
Hyunwoo Kim, Jack Hessel, Liwei Jiang, Peter West, Ximing Lu, Youngjae Yu, Pei Zhou, Ronan Bras, Malihe Alikhani, Gunhee Kim, Maarten Sap, Yejin Choi
| Challenge: | a dataset of 1.5 million conversations distilled from everyday spoken situations is limited in scale due to its associated costs. |
| Approach: | They propose to make SODA a publicly available, million-scale high-quality social dialogue dataset . they contextualize social commonsense knowledge from a knowledge graph to distill broad spectrum of social interactions . |
| Outcome: | The proposed dataset is the first publicly available, million-scale high-quality social dialogue dataset. |
Moral Stories: Situated Reasoning about Norms, Intents, Actions, and their Consequences (2021.emnlp-main)
Copied to clipboard
| Challenge: | aaron carroll: in social settings, human behavior is governed by unspoken rules of conduct rooted in societal norms . carroll and colleagues examine whether language generation models can serve as behavioral priors if they are not . they say we examine whether they can generate descriptions of actions that accomplish predefined goals . |
| Approach: | They propose to combine multiple expert models to improve quality of generated actions, consequences, and norms. |
| Outcome: | The proposed models significantly improve the quality of generated actions, consequences, and norms compared to baselines. |
Neural Theory-of-Mind? On the Limits of Social Intelligence in Large LMs (2022.emnlp-main)
Copied to clipboard
| Challenge: | We show that one of today’s largest language models lacks this kind of social intelligence out-of-the-box, using two tasks: SocialIQa and ToMi. |
| Approach: | They propose to use social intelligence and Theory of Mind to examine whether modern large-scale language models lack this kind of social intelligence out-of-the-box. |
| Outcome: | The proposed model lacks social intelligence out-of-the-box, and has well-below human accuracies on SocialIQa and ToMi, respectively. |
QuAC: Question Answering in Context (D18-1)
Copied to clipboard
Eunsol Choi, He He, Mohit Iyyer, Mark Yatskar, Wen-tau Yih, Yejin Choi, Percy Liang, Luke Zettlemoyer
| Challenge: | a dataset for Question Answering in Context contains 14K information-seeking QA dialogs . questions are often more open-ended, unanswerable, or only meaningful within the dialog context . |
| Approach: | They propose a dataset for Question Answering in Context that contains 14K dialogs . they use a student to ask questions about a Wikipedia section and a teacher to answer them . |
| Outcome: | The proposed dataset underperforms humans in a number of reference models . the dataset contains 14K information-seeking dialogs over sections from Wikipedia . |
ProsocialDialog: A Prosocial Backbone for Conversational Agents (2022.emnlp-main)
Copied to clipboard
Hyunwoo Kim, Youngjae Yu, Liwei Jiang, Ximing Lu, Daniel Khashabi, Gunhee Kim, Yejin Choi, Maarten Sap
| Challenge: | Existing dialogue systems fail to respond properly to potentially unsafe user utterances . existing systems either ignore or passively agree with unsafe content . |
| Approach: | They introduce a dataset to teach conversational agents to respond to problematic content following social norms. |
| Outcome: | The proposed dataset shows that ProsocialDialog generates more socially acceptable dialogues than existing models. |
Reading Books is Great, But Not if You Are Driving! Visually Grounded Reasoning about Defeasible Commonsense Norms (2023.emnlp-main)
Copied to clipboard
Seungju Han, Junhyeok Kim, Jack Hessel, Liwei Jiang, Jiwan Chung, Yejin Son, Yejin Choi, Youngjae Yu
| Challenge: | NormLens is a visual-grounded framework for understanding commonsense norms . state-of-the-art models are not well-aligned with human annotation, we show . |
| Approach: | They propose a visual-grounded framework to study commonsense norms by NormLens . they find that models are not well-aligned with human annotation . |
| Outcome: | The proposed model judgments and explanations are not well-aligned with human annotations. |
REV: Information-Theoretic Evaluation of Free-Text Rationales (2023.acl-long)
Copied to clipboard
| Challenge: | Existing metrics for rationale evaluation focus on the association between the rationale and a label, whereas REV is more sensitive to new information in free-text rationales. |
| Approach: | They propose a metric called REV to quantify the amount of new, label-relevant information in a rationale beyond the information already available in the input or the label. |
| Outcome: | The proposed metric is consistent with human judgments on rationale evaluations and provides more sensitive measurements of new information in free-text rationales. |
Back to the Future: Unsupervised Backprop-based Decoding for Counterfactual and Abductive Commonsense Reasoning (2020.emnlp-main)
Copied to clipboard
Lianhui Qin, Vered Shwartz, Peter West, Chandra Bhagavatula, Jena D. Hwang, Ronan Le Bras, Antoine Bosselut, Yejin Choi
| Challenge: | Existing methods for integrating past and future contexts are limited and require manual input. |
| Approach: | They propose an unsupervised decoding algorithm that incorporates past and future contexts using off-the-shelf, left-to-right language models and no supervision. |
| Outcome: | The proposed method outperforms unsupervised methods on abductive and counterfactual reasoning tasks. |
On-the-Fly Attention Modulation for Neural Generation (2021.findings-acl)
Copied to clipboard
Yue Dong, Chandra Bhagavatula, Ximing Lu, Jena D. Hwang, Antoine Bosselut, Jackie Chi Kit Cheung, Yejin Choi
| Challenge: | Degeneration of neural text is associated with insufficient learning of task-specific characteristics by the attention mechanism. |
| Approach: | They propose to use attention modulation to inject priors into inference to improve fluency, creativity, and commonsense reasoning in neural text generation models. |
| Outcome: | The proposed method improves fluency, creativity, and commonsense reasoning, and significantly reduces sentence-level repetition. |
Minding Language Models’ (Lack of) Theory of Mind: A Plug-and-Play Multi-Character Belief Tracker (2023.acl-long)
Copied to clipboard
| Challenge: | Empirical results show plug-and-play approach to reason about belief states of multiple characters in reading comprehension tasks is more precise and interpretable than previous approaches. |
| Approach: | They propose a plug-and-play approach to reason about the belief states of multiple characters in reading comprehension tasks via explicit symbolic representation. |
| Outcome: | The proposed algorithm improves theory of mind of off-the-shelf neural language models without supervision. |
TIMEDIAL: Temporal Commonsense Reasoning in Dialog (2021.acl-long)
Copied to clipboard
| Challenge: | Existing studies on pre-trained language models for dialog reasoning fail to understand context correctly. |
| Approach: | They propose to use a crowd-sourced English task and a time-based task to test models' temporal reasoning abilities in dialogs. |
| Outcome: | The proposed task and crowd-sourced English challenge set show that even the best performing models struggle on this task compared to humans. |
Symbolic Chain-of-Thought Distillation: Small Models Can Also “Think” Step-by-Step (2023.acl-long)
Copied to clipboard
| Challenge: | Symbolic Chain-of-thought Distillation (SCoTD) is a method to train a smaller student model on rationalizations sampled from a significantly larger teacher model. |
| Approach: | They propose a method to train a smaller student model on rationalizations from a larger teacher model. |
| Outcome: | The proposed method improves the performance of a student model in supervised and few-shot settings and especially for challenge sets. |
MacGyver: Are Large Language Models Creative Problem Solvers? (2024.naacl-long)
Copied to clipboard
Yufei Tian, Abhilasha Ravichander, Lianhui Qin, Ronan Le Bras, Raja Marjieh, Nanyun Peng, Yejin Choi, Thomas Griffiths, Faeze Brahman
| Challenge: | a new study examines the creative problem-solving capabilities of modern LLMs . it provides insight into the constrained problem- solving capabilities of both humans and AI . |
| Approach: | They use an automatically generated dataset to compare and contrast LLMs and humans to find out their creative problem-solving abilities. |
| Outcome: | The proposed dataset compares LLMs and humans in a constrained setting . it shows that humans excel in tasks they are familiar with but struggle with domain-specific knowledge . |
Making VLMs More Robot-Friendly: Self-Critical Distillation of Low-Level Procedural Reasoning (2025.emnlp-main)
Copied to clipboard
Chan Young Park, Jillian Fisher, Marius Memmel, Dipika Khullar, Seoho Yun, Abhishek Gupta, Yejin Choi
| Challenge: | Large language models (LLMs) are human-centric, but omit low-level, spatially grounded details needed for robotic execution. |
| Approach: | They propose a lightweight framework for vision-language procedural planning that enables iteratively critique, revise and verify their own plans without external supervision or teacher models. |
| Outcome: | a new framework outperforms weaker models 100X the size in vision-language procedural planning . the framework generates higher-quality, execution-ready plans that can be used at inference and fine-tuning . |
Reframing Instructional Prompts to GPTk’s Language (2022.findings-acl)
Copied to clipboard
| Challenge: | Using reframing techniques, we find that instructional prompts are easier to follow for Language Models (LMs) |
| Approach: | They propose reframing techniques for manual reformulation of prompts into more effective ones . they compare performance of LMs prompted with reframed instructions on 12 NLP tasks . |
| Outcome: | The reframing techniques used for prompt reformulation improve performance on 12 tasks . the techniques boost performance on LMs with different sizes compared with original prompts . |
Social Bias Frames: Reasoning about Social and Power Implications of Language (2020.acl-main)
Copied to clipboard
| Challenge: | Language has enormous power to project social biases and reinforce stereotypes on people. |
| Approach: | They propose a new conceptual formalism that aims to model the pragmatic frames in which people project social biases and power differentials onto others. |
| Outcome: | The proposed model can model the pragmatic frames in which people project social biases and power differentials onto others. |
UNcommonsense Reasoning: Abductive Reasoning about Uncommon Situations (2024.naacl-long)
Copied to clipboard
Wenting Zhao, Justin Chiu, Jena Hwang, Faeze Brahman, Jack Hessel, Sanjiban Choudhury, Yejin Choi, Xiang Li, Alane Suhr
| Challenge: | Existing work evaluating commonsense reasoning focuses on making inferences about common, everyday situations. |
| Approach: | They propose to use an English language corpus to investigate commonsense reasoning . they characterize performance differences between human explainers and best-performing large language models . |
| Outcome: | The proposed method reduces the loss rate of human-written explanations on commonsense reasoning compared with the vanilla supervised fine-tuning approach . |
Dataset Cartography: Mapping and Diagnosing Datasets with Training Dynamics (2020.emnlp-main)
Copied to clipboard
Swabha Swayamdipta, Roy Schwartz, Nicholas Lourie, Yizhong Wang, Hannaneh Hajishirzi, Noah A. Smith, Yejin Choi
| Challenge: | Large datasets have become commonplace in NLP research, but the emphasis on quantity has made it challenging to assess the quality of data. |
| Approach: | They propose a model-based tool to characterize and diagnose large datasets . they leverage the behavior of the model on individual instances during training . |
| Outcome: | Experiments on four datasets show that the tool can characterize and diagnose datasets with a model-based tool. |
Exposing the Limits of Video-Text Models through Contrast Sets (2022.naacl-main)
Copied to clipboard
| Challenge: | Recent video-text models can retrieve relevant videos based on text with high accuracy, but to what extent do they comprehend the semantics of the text? |
| Approach: | They propose a framework that probes video-text models with hard negatives . they leverage a pre-trained language model and a set of heuristics to create verb and person entity focused contrast sets. |
| Outcome: | The proposed framework erases the performance gap between CLIP-based methods and the earlier methods. |
PlotMachines: Outline-Conditioned Generation with Dynamic Plot State Tracking (2020.emnlp-main)
Copied to clipboard
| Challenge: | a novel task is to generate a coherent narrative consistent with an outline . large-scale language models are not sufficient in generating coherent narratives for the given outline despite their impressive generation performance . |
| Approach: | They propose a task of outline-conditioned story generation that generates a coherent narrative . they propose 'plotmachines' that tracks dynamic plot states and learns different writing styles . |
| Outcome: | The proposed model can generate a coherent story by tracking the dynamic plot states while conditioning on the input outline while generating the full story. |
From Dogwhistles to Bullhorns: Unveiling Coded Rhetoric with Language Models (2023.acl-long)
Copied to clipboard
| Challenge: | This work sheds light on the theoretical and applied importance of dogwhistles in both NLP and computational social science. |
| Approach: | They propose a typology of dogwhistles, curate a glossary of over 300 dogwhitles and analyze their usage in historical U.S. politicians’ speeches. |
| Outcome: | The proposed model identifies dogwhistles and their meanings and shows that harmful content containing dogwhitles avoids toxicity detection. |
Crystal: Introspective Reasoners Reinforced with Self-Feedback (2023.emnlp-main)
Copied to clipboard
| Challenge: | Existing knowledge-augmented reasoning methods fail to capture the *introspective* nature of knowledge required in commonsense reasoning. |
| Approach: | They propose a method to develop an introspective commonsense reasoner that introspects for knowledge statements related to the given question and makes an informed prediction. |
| Outcome: | The proposed method outperforms standard supervised finetuning and chain-of-thought distilled methods and enhances the transparency of the commonsense reasoning process. |
GENIE: Toward Reproducible and Standardized Human Evaluation for Text Generation (2022.emnlp-main)
Copied to clipboard
Daniel Khashabi, Gabriel Stanovsky, Jonathan Bragg, Nicholas Lourie, Jungo Kasai, Yejin Choi, Noah A. Smith, Daniel Weld
| Challenge: | Effective human evaluation of text generation tasks remains an important, open area for research. |
| Approach: | They propose a system for running standardized human evaluations across different generation tasks. |
| Outcome: | The proposed system produces standardized human evaluations across tasks . it crowdsources predictions and ranks systems on leaderboards . the proposed system is not reproducible over time and different annotator populations . |
Event2Mind: Commonsense Inference on Events, Intents, and Reactions (P18-1)
Copied to clipboard
| Challenge: | Using a crowdsourced corpus of 25,000 event phrases, we construct a new task that uses commonsense reasoning to reason about the likely intents and reactions of the event participants. |
| Approach: | They construct a crowdsourced corpus of 25,000 event phrases and use them to construct 'commonsense inference' they demonstrate that neural encoder-decoder models can compose embedding representations of previously unseen events and reason about the likely intents and reactions of the event participants. |
| Outcome: | The proposed task can be used to uncover implicit gender inequality in movie scripts. |
GO FIGURE: A Meta Evaluation of Factuality in Summarization (2021.findings-acl)
Copied to clipboard
| Challenge: | Recent advances in text generation systems produce fluent, coherent, relevant, and factually correct text. |
| Approach: | They propose a metaevaluation framework for evaluating factuality evaluation metrics . they propose five necessary conditions to evaluate factual metrics on diagnostic factuity data . |
| Outcome: | The proposed framework provides robust evaluation that is extensible to multiple types of factual consistency and standard generation metrics, including QA metrics. |
HALoGEN: Fantastic LLM Hallucinations and Where to Find Them (2025.acl-long)
Copied to clipboard
| Challenge: | generative large language models produce hallucinations that are not aligned with world knowledge or input context. |
| Approach: | They propose a hallucination benchmark framework that measures hallucinism in large language models . they evaluate 150,000 generations from 14 language models and find they are riddled with hallucinos . |
| Outcome: | The proposed framework evaluates 150,000 generations from 14 language models. |
StyleRemix: Interpretable Authorship Obfuscation via Distillation and Perturbation of Style Elements (2024.emnlp-main)
Copied to clipboard
| Challenge: | Authorship obfuscation methods that ignore author-specific stylistic features are often too rigid and lead to degradation of fluency and grammaticality. |
| Approach: | They propose an adaptive obfuscation method that perturbs stylistic elements of text . authors release a large set of 30K high-quality, long-form texts from a diverse set of 14 authors . |
| Outcome: | The proposed method outperforms state-of-the-art methods on an array of domains on automatic and human evaluation. |
When One LLM Drools, Multi-LLM Collaboration Rules (2026.acl-long)
Copied to clipboard
Shangbin Feng, Wenxuan Ding, Alisa Liu, Zifeng Wang, Weijia Shi, Yike Wang, Shannon Zejiang Shen, Xiaochuang Han, Hunter Lang, Chen-Yu Lee, Tomas Pfister, Yejin Choi, Yulia Tsvetkov
| Challenge: | a single general-purpose LLM is not enough to produce a reliable output, argues this paper . a multi-LLM collaboration approach addresses reliability, democratization, and pluralism . |
| Approach: | They argue that a single general-purpose LLM is not enough to produce a reliable output . they organize existing multi-LLM collaboration methods into a hierarchy based on access and information exchange . |
| Outcome: | The proposed method addresses reliability, democratization, and pluralism challenges a single LLM fails to produce a reliable output. |
Maieutic Prompting: Logically Consistent Reasoning with Recursive Explanations (2022.emnlp-main)
Copied to clipboard
Jaehun Jung, Lianhui Qin, Sean Welleck, Faeze Brahman, Chandra Bhagavatula, Ronan Le Bras, Yejin Choi
| Challenge: | Pre-trained language models struggle with consistent reasoning, and prompting methods are often noisy and inconsistent. |
| Approach: | They propose a few-shot inference method inspired by the Socratic way of conversation that generates a tree of explanations that bear logical relations between each other and frames it as a satisfiability problem. |
| Outcome: | The proposed method achieves 20% better accuracy than state-of-the-art prompting methods and performs competitively with supervised models. |
STEER: Unified Style Transfer with Expert Reinforcement (2023.findings-emnlp)
Copied to clipboard
| Challenge: | Experimental results show unified style transfer models outperform the 175B instruction-tuned GPT-3 on overall style transfer quality. |
| Approach: | They propose a unified style transfer framework that can transfer to multiple target styles from an arbitrary source style. |
| Outcome: | The proposed method outperforms the 175B instruction-tuned GPT-3 on overall style transfer quality despite being 226 times smaller in size . |
Perceptions to Beliefs: Exploring Precursory Inferences for Theory of Mind in Large Language Models (2024.emnlp-main)
Copied to clipboard
| Challenge: | While theory of mind (ToM) is naturally developed for humans in childhood, large language models (LLMs) exhibit inconsistency in ToM tasks, despite early reports of successful cases. |
| Approach: | They propose to evaluate human ToM precursors-perception inference and perception-to-belief inference-in large language models (LLMs) by annotating characters’ perceptions on ToMi and FANToM. |
| Outcome: | The proposed method significantly improves LLMs’ performance in false belief scenarios. |
Learning to Write with Cooperative Discriminators (P18-1)
Copied to clipboard
| Challenge: | Despite their local fluency, long-form text generated from RNNs is often generic, repetitive, and even self-contradictory. |
| Approach: | They propose a unified learning framework that can guide a base RNN generator towards more globally coherent generations by combining discriminators with a composite decoding objective. |
| Outcome: | The proposed framework can guide a base RNN generator towards more globally coherent generations by combining discriminators with the base RRN generator through a composite decoding objective. |
RewardBench: Evaluating Reward Models for Language Modeling (2025.findings-naacl)
Copied to clipboard
Nathan Lambert, Valentina Pyatkin, Jacob Morrison, Lester James Validad Miranda, Bill Yuchen Lin, Khyathi Chandu, Nouha Dziri, Sachin Kumar, Tom Zick, Yejin Choi, Noah A. Smith, Hannaneh Hajishirzi
| Challenge: | Evaluating reward models presents an opportunity to understand the opaque technologies used for alignment of language models. |
| Approach: | They present a benchmark dataset and code-base for evaluation of reward models . they use prompt-chosen-rejected trios to benchmark how they perform on queries . |
| Outcome: | The proposed dataset compares RMs with other models on a set of questions. |
Annotators with Attitudes: How Annotator Beliefs And Identities Bias Toxic Language Detection (2022.naacl-main)
Copied to clipboard
| Challenge: | toxicity annotations are often ignored because of its subjective nature and lack of nuance. |
| Approach: | They examine the effect of annotator identities and beliefs on toxic language annotations by considering posts with three characteristics: anti-Black language, African American English (AAE) dialect, and vulgarity. |
| Outcome: | The findings show strong associations between annotator identity and beliefs and ratings of toxicity. |
LOTUS: A Leaderboard for Detailed Image Captioning from Quality to Societal Bias and User Preferences (2025.acl-industry)
Copied to clipboard
Yusuke Hirota, Boyi Li, Ryo Hachiuma, Yueh-Hua Wu, Boris Ivanovic, Marco Pavone, Yejin Choi, Yu-Chiang Frank Wang, Yuta Nakashima, Chao-Han Huck Yang
| Challenge: | Large Vision-Language Models (LVLMs) have transformed image captioning . existing evaluations lack standardized criteria and a standardized evaluation framework . |
| Approach: | They propose a leaderboard for evaluating detailed captions that addresses three main gaps in existing evaluations: lack of standardized criteria, bias-aware assessments, and user preference considerations. |
| Outcome: | The proposed model evaluates caption quality, descriptiveness, risks, and societal biases while tailoring criteria to user preferences. |
Benchmarking Hierarchical Script Knowledge (N19-1)
Copied to clipboard
| Challenge: | Understanding procedural language requires reasoning about hierarchical and temporal relations between events. |
| Approach: | They propose a hierarchical script learning dataset and a cloze task to match video captions with missing procedural details. |
| Outcome: | The proposed model matches video captions with missing procedural details to find out if they can understand the language. |
Is GPT-3 Text Indistinguishable from Human Text? Scarecrow: A Framework for Scrutinizing Machine Text (2022.acl-long)
Copied to clipboard
| Challenge: | a recent study has reported that crowdsourcing cannot distinguish between machine-authored and human-authored text. |
| Approach: | They propose a framework called Scarecrow for scrutinizing machine text via crowd annotation . they use crowd annotation to identify redundancy, commonsense errors, and incoherence . |
| Outcome: | The proposed method quantifies gaps between human-authored and machine-generated text . it can detect redundancy, commonsense errors, and incoherence . |
We’re Afraid Language Models Aren’t Modeling Ambiguity (2023.emnlp-main)
Copied to clipboard
Alisa Liu, Zhaofeng Wu, Julian Michael, Alane Suhr, Peter West, Alexander Koller, Swabha Swayamdipta, Noah Smith, Yejin Choi
| Challenge: | Ambiguity is an intrinsic feature of natural language, allowing us to anticipate misunderstandings and revise our interpretations as listeners. |
| Approach: | They use AmbiEnt to capture ambiguity in a sentence and analyze it to evaluate pretrained LMs. |
| Outcome: | The proposed model can flag political claims in the wild that are misleading due to ambiguity. |
Sounding Board: A User-Centric and Content-Driven Social Chatbot (N18-5)
Copied to clipboard
Hao Fang, Hao Cheng, Maarten Sap, Elizabeth Clark, Ari Holtzman, Yejin Choi, Noah A. Smith, Mari Ostendorf
| Challenge: | Sounding Board is a social chatbot that can hold a coherent conversation with humans . the system is user-centric in that users can control the topic of conversation, while the system adapts to the user's needs. |
| Approach: | They present Sounding Board, a social chatbot that won the 2017 Amazon Alexa Prize. |
| Outcome: | The system is user-centric in that users can control the topic of conversation, while the system adapts to the user's needs. |
Neural Metaphor Detection in Context (D18-1)
Copied to clipboard
| Challenge: | Existing models focus on limited forms of linguistic context, such as unigrams. |
| Approach: | They propose end-to-end neural models for detecting metaphorical word use in context . they show that bi-directional biLSTM models which operate on complete sentences work well . |
| Outcome: | The proposed models show that they can learn rich contextual word representations . they are compared to previous models which focused on limited linguistic context . |
Rainier: Reinforced Knowledge Introspector for Commonsense Question Answering (2022.emnlp-main)
Copied to clipboard
| Challenge: | Recent research shows that relevant knowledge can provide useful context for commonsense tasks. |
| Approach: | They propose a method that learns to generate contextually relevant knowledge in response to given questions. |
| Outcome: | The proposed method shows consistent gains over 9 commonsense benchmarks. |
Modular Transformers: Compressing Transformers into Modularized Layers for Flexible Efficient Inference (2023.findings-acl)
Copied to clipboard
| Challenge: | Pre-trained sequence-to-sequence models have advanced the state of the art on text generation tasks. |
| Approach: | They introduce a modular encoder-decoder framework for flexible sequence-to-sequence model compression. |
| Outcome: | The proposed framework can achieve flexible compression ratios from 1.1x to 6x with little to moderate relative performance drop. |
Information-Guided Identification of Training Data Imprint in (Proprietary) Large Language Models (2025.naacl-long)
Copied to clipboard
Abhilasha Ravichander, Jillian Fisher, Taylor Sorensen, Ximing Lu, Maria Antoniak, Bill Yuchen Lin, Niloofar Mireshghallah, Chandra Bhagavatula, Yejin Choi
| Challenge: | Lack of transparency in training data is limiting external oversight and inspection of LLMs for issues such as copyright infringement and data contamination. |
| Approach: | They propose a method to identify training data known to proprietary LLMs without requiring access to model weights or token probabilities by using information-guided probes. |
| Outcome: | The proposed method can identify training data known to proprietary LLMs without access to model weights or token probabilities. |
Edited Media Understanding Frames: Reasoning About the Intent and Implications of Visual Misinformation (2021.acl-long)
Copied to clipboard
| Challenge: | Edited media frames are structured annotations with respect to intents, emotional reactions, attacks on individuals, and the implications of disinformation. |
| Approach: | They propose a new formalism to understand visual media manipulation as structured annotations with respect to intents, emotional reactions, attacks on individuals, and the implications of disinformation. |
| Outcome: | The proposed model obtains promising results on a dataset with 56k question-answer pairs written in rich natural language. |
Deep Communicating Agents for Abstractive Summarization (N18-1)
Copied to clipboard
| Challenge: | Empirical results show that multiple communicating agents produce a better summary than extractive summarization. |
| Approach: | They propose an encoder-decoder architecture that uses deep communicating agents to represent a long document for abstractive summarization. |
| Outcome: | Empirical results show that multiple communicating agents produce a better summary than baselines. |
In Search of the Long-Tail: Systematic Generation of Long-Tail Inferential Knowledge via Logical Rule Guided Search (2024.emnlp-main)
Copied to clipboard
Huihan Li, Yuting Ning, Zeyi Liao, Siyuan Wang, Xiang Li, Ximing Lu, Wenting Zhao, Faeze Brahman, Yejin Choi, Xiang Ren
| Challenge: | Logic-Induced-Knowledge-Search (LINK) is a framework for generating factually-correct yet long-tail inferential knowledge. |
| Approach: | They introduce a framework to obtain factually-correct yet long-tail inferential statements using variable-wise prompting grounded on symbolic rules. |
| Outcome: | The proposed framework is able to obtain factually-correct yet long-tail inferential statements while ensuring factual correctness. |
Surface Form Competition: Why the Highest Probability Answer Isn’t Always Right (2021.emnlp-main)
Copied to clipboard
| Challenge: | Large language models have shown promising results in zero-shot settings due to surface form competition . since probability mass is finite, this lowers the probability of the correct answer . |
| Approach: | They propose a scoring function that compensates for surface form competition by reweighing each option according to its a priori likelihood. |
| Outcome: | The proposed scoring function achieves consistent gains in zero-shot over calibrated and uncalibrated scoring functions on all GPT-2 and GPT-3 models on a variety of multiple choice datasets. |
Vera: A General-Purpose Plausibility Estimation Model for Commonsense Statements (2023.emnlp-main)
Copied to clipboard
| Challenge: | Despite the advances of language models, they still produce text that contains trivial commonsense errors. |
| Approach: | They propose a general-purpose commonsense statement verification model that learns to estimate the plausibility of declarative statements based on commonsensical knowledge. |
| Outcome: | The proposed model outperforms existing models that can be repurposed for commonsense verification, even including GPT-3.5/ChatGPT/GPT-4. |
Can LLMs Reason with Rules? Logic Scaffolding for Stress-Testing and Improving LLMs (2024.acl-long)
Copied to clipboard
| Challenge: | Large language models (LLMs) have impressive human-like performance across various reasoning tasks, but their mastery of underlying inferential rules falls short of human capabilities. |
| Approach: | They propose a logic scaffolding inferential rule generation framework to construct an infer- ential rule base, ULogic, comprising both primitive and compositional rules across five domains. |
| Outcome: | The proposed model improves the ability to generate accurate, complex and abstract conclusions and premises and improves various commonsense reasoning tasks. |
ClarifyDelphi: Reinforced Clarification Questions with Defeasibility Rewards for Social and Moral Situations (2023.acl-long)
Copied to clipboard
Valentina Pyatkin, Jena D. Hwang, Vivek Srikumar, Ximing Lu, Liwei Jiang, Yejin Choi, Chandra Bhagavatula
| Challenge: | Changing contexts can flip the moral judgment of an action. |
| Approach: | They propose an interactive system that learns to ask clarification questions to elicit salient contexts of a social or moral situation. |
| Outcome: | The proposed system generates more relevant, informative and defeasible questions compared to baselines. |
Referee: Reference-Free Sentence Summarization with Sharper Controllability through Symbolic Knowledge Distillation (2022.emnlp-main)
Copied to clipboard
| Challenge: | a new framework for sentence summarization is available that can be trained reference-free . a high-quality dataset of sentence-summary pairs with varying degrees of compression ratios is obtained . |
| Approach: | They propose a framework for sentence summarization that can be trained reference-free . they propose 'referee' that iteratively distills latent knowledge into better models . |
| Outcome: | The proposed framework outperforms existing models in the use of explicit examples from teacher models without compromising the quality of the summarization. |
Robust Navigation with Language Pretraining and Stochastic Sampling (D19-1)
Copied to clipboard
Xiujun Li, Chunyuan Li, Qiaolin Xia, Yonatan Bisk, Asli Celikyilmaz, Jianfeng Gao, Noah A. Smith, Yejin Choi
| Challenge: | Existing methods to learn visual representations and action decoding schemes are limited to previously unseen instructions and environments. |
| Approach: | They propose a stochastic sampling scheme to reduce the gap between the expert actions in training and sampled actions in test to correct its own mistakes. |
| Outcome: | The proposed methods achieve 6% absolute gain over the previous best results on the Room-to-Room benchmark. |
Reframing Human-AI Collaboration for Generating Free-Text Explanations (2022.naacl-main)
Copied to clipboard
| Challenge: | Large language models are capable of generating fluent-appearing text with little task-specific supervision. |
| Approach: | They propose a pipeline that combines GPT-3 with a supervised filter that incorporates binary acceptability judgments from humans in the loop. |
| Outcome: | The proposed model can generate freetext explanations in a fewshot setting with human-written examples. |
Symbolic Knowledge Distillation: from General Language Models to Commonsense Models (2022.naacl-main)
Copied to clipboard
Peter West, Chandra Bhagavatula, Jack Hessel, Jena Hwang, Liwei Jiang, Ronan Le Bras, Ximing Lu, Sean Welleck, Yejin Choi
| Challenge: | Prior studies suggested pre-trained language models possess limited understanding of commonsense knowledge despite otherwise stellar performance on leaderboards. |
| Approach: | They propose a framework that uses larger models to teach smaller models by distilling knowledge symbolically as text in addition to the neural model. |
| Outcome: | The proposed framework is based on a general language model teacher's commonsense knowledge graphs and a neural commonsensing model surpassing the teacher model's in all three criteria. |
Zero-shot Multimodal Document Retrieval via Cross-modal Question Generation (2025.emnlp-main)
Copied to clipboard
| Challenge: | Existing multimodal large language models struggle when faced with unseen domains or languages. |
| Approach: | They propose a framework that leverages the broad knowledge of an MLLM to generate cross-modal pre-questions (preQs) before retrieval. |
| Outcome: | Experiments show that PREMIR outperforms existing retrievers on out-of-distribution benchmarks, including closed-domain and multilingual settings, outperforming strong baselines across all metrics. |
Faking Fake News for Real Fake News Detection: Propaganda-Loaded Training Data Generation (2023.acl-long)
Copied to clipboard
| Challenge: | despite advances in detecting fake news, there is a sizable gap between machine-generated and human-authored fake news . a nave solution is to collect human-written news articles that contain inaccurate information by crawling untrustworthy news media. |
| Approach: | They propose a framework for generating training examples informed by the styles and strategies of human-authored propaganda. |
| Outcome: | The proposed framework improves detection of human-written disinformation by 3.62–7.69% on two public datasets. |
NeuroLogic A*esque Decoding: Constrained Text Generation with Lookahead Heuristics (2022.naacl-main)
Copied to clipboard
Ximing Lu, Sean Welleck, Peter West, Liwei Jiang, Jungo Kasai, Daniel Khashabi, Ronan Le Bras, Lianhui Qin, Youngjae Yu, Rowan Zellers, Noah A. Smith, Yejin Choi
| Challenge: | Existing paradigms for text generation are left-to-right decoding from autoregressive language models. |
| Approach: | They propose a decoding algorithm that incorporates heuristic estimates of future cost that are efficient for large-scale language models. |
| Outcome: | The proposed method outperforms baselines on five generation tasks and achieves new state-of-the-art performance on table-to-text generation, constrained machine translation, and keyword-constrained generation. |