Papers by Zhen Guo

21 papers
SciPedia: Unlocking the Value of Scientific Data for Pre-training (2026.acl-long)

Copied to clipboard

Challenge: High-quality scientific data is critical for advancing LLMs, yet academic literature remains underutilized.
Approach: They construct a large-scale raw scientific corpus but identify a critical Learnability Gap . they develop a multi-stage pipeline featuring content cleaning and pedagogical augmentation .
Outcome: The proposed approach boosts average performance by +2.12 (3B) and +2.95 (7B) on in-domain tasks.
FC-Attack: Jailbreaking Multimodal Large Language Models via Auto-Generated Flowcharts (2025.findings-emnlp)

Copied to clipboard

Challenge: Recent research shows that multimodal large language models are vulnerable to jailbreak attacks .
Approach: They propose a jailbreak attack method based on auto-generated flowcharts . the flowchartings are then combined with a benign textual prompt to execute the attack .
Outcome: The proposed method achieves an attack success rate of up to 96% via images and 78% via videos across multiple MLLMs.
Temporal Token Matters: Investigating and Interpreting the Consistency of Temporal Ordering in Large Language Models (2026.findings-acl)

Copied to clipboard

Challenge: Large Language Models (LLMs) exhibit notable deficiencies in temporal reasoning . phrasing changes can lead LLMs to produce inconsistent outputs .
Approach: They investigate the mechanistic interpretability of temporal ordering within event temporal reasoning . they identify a sparse subset of attention heads that are causally responsible for reasoning outcomes .
Outcome: The proposed model outperforms other models in a variety of tasks and is validated by intervention-based experiments.
Debias NLU Datasets via Training-free Perturbations (2023.findings-emnlp)

Copied to clipboard

Challenge: Existing approaches to debiase NLU models capture biased features that are independent of the task but spuriously correlated to labels.
Approach: They propose a framework that conducts training-free perturbations on samples containing biased features to Debias NLU Datasets.
Outcome: The proposed framework shows competitive performance with previous state-of-the-art debiasing strategies.
Beyond Verbal Cues: Emotional Contagion Graph Network for Causal Emotion Entailment (2025.findings-acl)

Copied to clipboard

Challenge: Recent studies have focused on identifying the causes of emotions by understanding verbal contextual utterances, but this study often lacks recognizing the underlying emotional stimuli present in these utterrances.
Approach: They propose an Emotional Contagion Graph Network that simulates the impact of non-verbal emotional cues on the counterpart’s emotions.
Outcome: The proposed model is compared with state-of-the-art models on a benchmark dataset and the results are encouraging.
Noise-Robust Fine-Tuning of Pretrained Language Models via External Guidance (2023.findings-emnlp)

Copied to clipboard

Challenge: Pretrained Language Models (PLMs) are advanced but data labels are noisy due to the complex annotation process.
Approach: They propose a framework for fine-tuning PLMs using noisy labels that incorporates guidance from Large Language Models like ChatGPT.
Outcome: Experiments on synthetic and real-world noisy datasets show that the proposed framework outperforms the state-of-the-art framework.
ToolPRMBench: Evaluating and Advancing Process Reward Models for Tool-using Agents (2026.findings-acl)

Copied to clipboard

Challenge: Reward-guided search methods have shown potential in enhancing tool-using agents . however, there is a lack of reliable evaluation benchmarks for PRMs in tool-use settings .
Approach: They propose a large-scale benchmark specifically designed to evaluate PRMs for tool-using agents.
Outcome: The proposed benchmark shows that tool reward models perform better in tool-using environments.
PLATO-XL: Exploring the Large-scale Pre-training of Dialogue Generation (2022.findings-aacl)

Copied to clipboard

Challenge: Experimental results show PLATO-XL achieves state-of-the-art results across multiple conversational tasks.
Approach: They propose to train PLATO-XL models with up to 11 billion parameters, trained on Chinese and English social media conversations.
Outcome: The proposed model achieves state-of-the-art on multiple conversational tasks, verifying its potential as a foundation model of conversational AI.
Proactive Human-Machine Conversation with Explicit Conversation Goal (P19-1)

Copied to clipboard

Challenge: Typical human-machine conversation systems only use utterances and responses as training data, which results in uninformative and inappropriate responses.
Approach: They propose a dataset where one acts as a conversation leader and the other as 'follower' they establish baseline results on a 270K utterances and 30k dialogues dataset using state-of-the-art models.
Outcome: The proposed model can generate diverse multi-turn conversations using knowledge from a new dataset .
XAL: EXplainable Active Learning Makes Classifiers Better Low-resource Learners (2024.naacl-long)

Copied to clipboard

Challenge: Existing methods for active learning rely on model uncertainty or disagreement to pick unlabeled data, leading to over-confidence in superficial patterns and lack of exploration.
Approach: They propose to use a bi-directional encoder and a uni-directional decoder to generate and score an explanation for low-resource text classification.
Outcome: The proposed model improves on 9 strong baselines on six datasets and can generate explanations for its predictions.
Boosting Policy and Process Reward Models with Monte Carlo Tree Search in Open-Domain QA (2025.findings-acl)

Copied to clipboard

Challenge: Experimental results show that our approach can effectively improve the performance of both the policy model and the reward model.
Approach: They propose to use Monte Carlo Tree Search for both policy model improvement and reward model improvement to bridge it to more subtle open-domain question answering.
Outcome: The proposed approach surpasses existing methods for annotation and training data with fewer data points and achieves better performance in test-time scaling strategies.
LLMRouterBench: A Massive Benchmark and Unified Framework for LLM Routing (2026.findings-acl)

Copied to clipboard

Challenge: Large language model (LLM) routing assigns each query to the best suitable model from an ensemble.
Approach: They introduce a large-scale benchmark and unified framework for LLM routing . they find that many routing methods exhibit similar performance under unified evaluation .
Outcome: The proposed benchmark provides comprehensive metrics for both performance-oriented and performance-cost trade-off routing.
CDConv: A Benchmark for Contradiction Detection in Chinese Conversations (2022.emnlp-main)

Copied to clipboard

Challenge: Existing methods for detecting dialogue contradictions are difficult due to contextualization nature of conversations.
Approach: They propose a benchmark for Contradiction Detection in Chinese Conversations . they use automatic conversation generation to simulate common user behaviors .
Outcome: The proposed benchmark simulated the user behaviors that trigger chatbots to make contradictions . the results show that the current state-of-the-art chatbot can be easily goaded into making contradictions.
How Do LLMs and VLMs Understand Viewpoint Rotation Without Vision? An Interpretability Study (2026.acl-long)

Copied to clipboard

Challenge: Existing studies on spatial intelligence from the perspective of visual-spatial intelligence have not explored whether visual intelligence alone is sufficient to endow models with spatial intelligence.
Approach: They propose to use a linguistic perspective to investigate spatial intelligence from a theoretical perspective.
Outcome: The proposed model performs poorly on the proposed dataset while human can easily achieve 100% accuracy.
AutoFigure-Edit: Generating Editable Scientific Illustrations via Reference-Guided Styling (2026.acl-demo)

Copied to clipboard

Challenge: Existing automated systems for scientific illustrations are limited in editability, stylistic controllability, and efficiency.
Approach: They propose an end-to-end system that generates fully editable scientific illustrations from long-form scientific text while enabling flexible style adaptation through user-provided reference images.
Outcome: The proposed system generates fully editable scientific illustrations from long-form scientific texts while enabling flexible style adaptation through user-provided reference images.
PyramidCodec: Hierarchical Codec for Long-form Music Generation in Audio Domain (2024.findings-emnlp)

Copied to clipboard

Challenge: Existing approaches to generate long music are inefficient and lack of structured representation.
Approach: They propose a hierarchical discrete representation of audio for long audio-domain music generation using residual vector quantization on different levels of features.
Outcome: The proposed method achieves competitive performance in terms of reconstruction quality and token per second (TPS) the proposed method facilitates training a language model that can generate well-structured long-form music for up to 3 minutes.
Hierarchical Reinforcement Learning with Augmented Step-Level Transitions for LLM Agents (2026.acl-long)

Copied to clipboard

Challenge: Existing Large language model agents rely on increasingly long interaction histories, resulting in high computational cost and limited scalability.
Approach: They propose a hierarchical reinforcement learning framework that enables step-level learning by conditioning only on single-step transitions rather than full interaction histories.
Outcome: The proposed framework outperforms baselines on ScienceWorld and ALFWorld benchmarks in terms of performance and generalization while reducing token usage.
Emotion-Anchored Contrastive Learning Framework for Emotion Recognition in Conversation (2024.findings-naacl)

Copied to clipboard

Challenge: Emotion Recognition in Conversation (ERC) is a task that aims to identify the emotions behind each utterance in a conversation.
Approach: They propose an Emotion-Anchored Contrastive Learning framework that generates more distinguishable utterance representations for similar emotions.
Outcome: The proposed framework achieves state-of-the-art on similar emotions and performs well on similar ones.
On-Policy Self-Distillation for Efficient Diffusion Language Models with Early-Stage Calibration (2026.findings-acl)

Copied to clipboard

Challenge: Recent studies have demonstrated that masked diffusion models (MDMs) can surpass autoregressive models (ARMs) in various tasks.
Approach: They propose a method to calibrate early token predictions without demonstration data by distilling an unnormalized target distribution into the original model.
Outcome: Experiments on math, planning, and RLHF tasks show that COPSD improves both effectiveness and efficiency, and further enhances performance when combined with supervised fine-tuning.
Can LLMs See Without Pixels? Benchmarking Spatial Intelligence from Textual Descriptions (2026.findings-acl)

Copied to clipboard

Challenge: Existing advances in Spatial Intelligence rely on vision-Language Models . however, a critical question remains: does spatial understanding originate from visual encoders?
Approach: They propose to evaluate the SI performance of Large Language Models without pixel-level input.
Outcome: The proposed benchmark challenges large language models to perform symbolic reasoning rather than visual pattern matching.
PLATO-2: Towards Building an Open-Domain Chatbot via Curriculum Learning (2021.findings-acl)

Copied to clipboard

Challenge: PLATO-2 is a high-quality open-domain chatbot that can generate one-to-many mappings and improve response quality.
Approach: They propose a curriculum learning process to build a high-quality open-domain chatbot . they use a coarse-grained generation model and latent variables to train a generative model .
Outcome: The proposed model improves on Chinese and English data and can generate diverse responses and select the best response.

What is GenGO?

GenGO is an NLP powered publication search system. It currenctly indexes 30k+ papers from ACL Anthology, and implements multi-aspect summarization, semantic search, and more!

Information

About
Limitations