Papers by Madian Khabsa

18 papers
SMARTAVE: Structured Multimodal Transformer for Product Attribute Value Extraction (2022.findings-emnlp)

Copied to clipboard

Challenge: Existing methods for product attribute value extraction are noisy and incomplete with missing values for most retailers.
Approach: They propose a Structure Mltimodal trAnsformeR for producT Attribute Value Extraction which jointly encodes the structured product information from multiple modalities.
Outcome: The proposed method outperforms state-of-the-art methods on two multimodal product datasets.
MART: Improving LLM Safety with Multi-round Automatic Red-Teaming (2024.naacl-long)

Copied to clipboard

Challenge: Existing red-teaming methods for large language models often discover safety risks without addressing them.
Approach: They propose a multi-round automatic red-teaming method that incorporates both adversarial prompt writing and safe response generation.
Outcome: The proposed method significantly increases red-teaming scalability and the safety of the target LLM.
Beyond Reasoning Gains: Mitigating General-Capability Forgetting in Large Reasoning Models (2026.findings-acl)

Copied to clipboard

Challenge: Reinforcement learning with verifiable rewards (RLVR) has delivered impressive gains in mathematical and multimodal reasoning . however, the recipe introduces a significant risk of capability regression, where models forget foundational skills after prolonged training without employing regularization strategies.
Approach: They propose a replay strategy with dynamic objective reweighting for general knowledge preservation using short-horizon signals of convergence and instability.
Outcome: The proposed method preserves general capabilities and improves reasoning . it can be applied to existing RLVR pipelines without training additional models or tuning .
Residual Prompt Tuning: improving prompt tuning with residual reparameterization (2023.findings-acl)

Copied to clipboard

Challenge: Prompt tuning is one of the most parameter-efficient approaches for parameter-effective tuning of pre-trained language models.
Approach: They propose to reparameterize soft prompt embeddings using a shallow network with a residual connection and use it to tune prompt embeds P.
Outcome: The proposed method outperforms prompt tuning on SuperGLUE, T5-Base and BERT-Bass models and can reduce the prompt length by 10 times without hurting performance.
MUSTIE: Multimodal Structural Transformer for Web Information Extraction (2023.acl-long)

Copied to clipboard

Challenge: Recent sequential modeling approaches focus on extracting information from textual sources while ignoring rich information from other modalities such as image and web layout.
Approach: They propose a novel MUltimodal Structural Transformer that integrates multiple modalities for web information extraction.
Outcome: The proposed model outperforms existing methods on WebSRC and Common Crawl benchmarks.
Generating Hashtags for Short-form Videos with Guided Signals (2023.acl-long)

Copied to clipboard

Challenge: Short-form video hashtag recommendation (SVHR) is a classification or ranking problem that selects hashtags from a set of limited candidates.
Approach: They propose a short-form video hashtag recommendation task that better represents how hashtags are created naturally by retrieving relevant hashtags from a large-scale hashtag pool as extra guidance signals.
Outcome: The proposed model outperforms strong classification baselines on two short-form video datasets and the guidance signals boost the performance by 8.11 and 2.17 absolute ROUGE-1 scores on average.
XLM-V: Overcoming the Vocabulary Bottleneck in Multilingual Masked Language Models (2023.emnlp-main)

Copied to clipboard

Challenge: Large multilingual models rely on a single vocabulary shared across 100+ languages . this vocabulary bottleneck limits the representational capabilities of multilingual model XLM-R .
Approach: They propose a new approach for scaling to large multilingual vocabularies by de-emphasizing token sharing between languages with little lexical overlap and assigning vocabulary capacity to achieve sufficient coverage for each individual language.
Outcome: The proposed model outperforms XLM-R on all language tasks and is particularly effective on low-resource tasks.
Sparse Distillation: Speeding Up Text Classification by Using Bigger Student Models (2022.naacl-main)

Copied to clipboard

Challenge: Existing methods to reduce inference cost by distilling transformer models into lightweight student models are limited for high-volume use cases.
Approach: They propose to distill state-of-the-art transformer models into lightweight student models to reduce computation cost at inference time.
Outcome: The proposed pipeline achieves up to 600x speed-up on GPUs and CPUs on six single-sentence text classification tasks and in domain generalization settings.
Quantifying Adaptability in Pre-trained Language Models with 500 Tasks (2022.naacl-main)

Copied to clipboard

Challenge: a recent study examines the features and limits of LM adaptability to new tasks . many questions about the nature and limits remain unanswered .
Approach: They evaluate adaptability to new tasks using a new benchmark, TaskBench500 . they find adaptation procedures differ dramatically in their ability to memorize small datasets .
Outcome: The proposed benchmark compares 500 procedurally generated sequence modeling tasks to a new benchmark.
The Belebele Benchmark: a Parallel Reading Comprehension Dataset in 122 Language Variants (2024.acl-long)

Copied to clipboard

Challenge: Existing benchmarks for text comprehension only cover 30 languages, but lack of labeled data is a major obstacle to building functional systems in most languages.
Approach: They present a multiple-choice machine reading comprehension dataset spanning 122 languages . they use it to evaluate the capabilities of multilingual masked language models and large language models .
Outcome: The proposed dataset enables the evaluation of text models in high-, medium- and low-resource languages.
MixPAVE: Mix-Prompt Tuning for Few-shot Product Attribute Value Extraction (2023.findings-acl)

Copied to clipboard

Challenge: Existing methods for product attribute value extraction focus on extracting values for a set of known attributes with sufficient training data.
Approach: They propose a prompt tuning approach to extract attributes from product information using mixed prompts.
Outcome: The proposed approach improves on two product benchmarks and shows parameter-efficient training and avoids model overfitting.
UniPELT: A Unified Framework for Parameter-Efficient Language Model Tuning (2022.acl-long)

Copied to clipboard

Challenge: Existing methods for parameter-efficient language model tuning (PELT) match the performance of fine-tuning with fewer trainable parameters.
Approach: They propose a framework which integrates different PELT methods as submodules and learns to activate the ones that best suit the current data or task setup via gating mechanism.
Outcome: The proposed framework outperforms fine-tuning methods on the GLUE benchmark and achieves 14% gains over the best individual PELT method.
On the Influence of Masking Policies in Intermediate Pre-training (2021.emnlp-main)

Copied to clipboard

Challenge: Existing studies show that inserting an intermediate pre-training stage improves performance of masked language models.
Approach: They propose methods to automate the discovery of optimal masking policies via direct supervision or meta-learning.
Outcome: The proposed method outperforms the heuristic of masking named entities on TriviaQA and can be generalizable beyond that task.
To Pretrain or Not to Pretrain: Examining the Benefits of Pretrainng on Resource Rich Tasks (2020.acl-main)

Copied to clipboard

Challenge: Existing studies on pretraining NLP models with variants of Masked Language Model (MLM) objectives have shown that the number of training samples used in the downstream task is limited.
Approach: They propose to use MLM objectives to pretrain NLP models with variants of Masked Language Model (MLM) objectives to improve accuracy on downstream tasks.
Outcome: The proposed model can reach a diminishing return point as the supervised data size increases significantly.
Keeping Notes: Conditional Natural Language Generation with a Scratchpad Encoder (P19-1)

Copied to clipboard

Challenge: Qualitative assessments in the form of human judgements (question generation), attention visualization (MT), and sample output (summarization) provide further evidence of the ability of Scratchpad to generate fluent and expressive output.
Approach: They propose to use the encoder as a "scratchpad" memory to keep track of what has been generated and guide future generation.
Outcome: The proposed mechanism improves the fluency of seq2seq models on three well-studied natural language generation tasks.
Effective Long-Context Scaling of Foundation Models (2024.naacl-long)

Copied to clipboard

Challenge: Large language models (LLMs) are rapidly deployed and continue to evolve through scaling.
Approach: They propose a method to train strong long-context LLMs that are capable of utilizing massive context windows of up to 32,000 tokens.
Outcome: The proposed model can surpass gpt-3.5-turbo-16k's overall performance on long-context benchmarks with a cost-effective instruction tuning procedure that is free of expensive annotations.
RoAST: Robustifying Language Models via Adversarial Perturbation with Selective Training (2023.findings-emnlp)

Copied to clipboard

Challenge: Several perspectives of robustness for pre-trained language models have been studied independently, but lacking a unified consideration in multiple perspectives.
Approach: They propose a technique to enhance the multi-perspective robustness of LMs by introducing adversarial perturbation while the model parameters are selectively updated upon their relative importance.
Outcome: The proposed technique improves the robustness of LMs by incorporating four perspectives on model robustness.
On Unifying Misinformation Detection (2021.naacl-main)

Copied to clipboard

Challenge: On any given day, 2.5 quintillion bytes of information are created on the Internet, a figure that is only expected to increase in the coming years.
Approach: They propose a general-purpose misinformation model that jointly models multiple domains of misinformation with a single, unified setup.
Outcome: The proposed model is useful for few-shot learning of unseen misinformation tasks/datasets and generalizability to unseense events.

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