Papers by Lambert Mathias
VideoMind: Thinking in Steps for Long Video Understanding (2026.eacl-industry)
Copied to clipboard
Shubhang Bhatnagar, Renxiong Wang, Kapil Krishnakumar, Adel Ahmadyan, Zhaojiang Lin, Lambert Mathias, Xin Luna Dong, Babak Damavandi, Narendra Ahuja, Seungwhan Moon
| Challenge: | Multimodal Large Language Models struggle with Long Video Understanding due to their limited context window and the distributed nature of salient information across many redundant frames. |
| Approach: | They propose a training framework that mimics a human reasoning process to train Long Video Understanding models. |
| Outcome: | The proposed framework achieves 77.6% performance on Video MME, LongVideo, and MLVU benchmarks while yielding 5% improvement on Llama 4 Scout. |
The Alexa Meaning Representation Language (N18-3)
Copied to clipboard
Thomas Kollar, Danielle Berry, Lauren Stuart, Karolina Owczarzak, Tagyoung Chung, Lambert Mathias, Michael Kayser, Bradford Snow, Spyros Matsoukas
| Challenge: | a new meaning representation language for spoken language is introduced for Alexa . AMRL provides a common representation for how people communicate in spoken language . there is no mechanism to represent ambiguity, forcing the choice of a fixed interpretation for ambiguous utterances. |
| Approach: | They introduce a meaning representation for spoken language, the Alexa meaning represent language . they use a spoken language dataset to collect a sample of utterances from eight domains . |
| Outcome: | The proposed representation provides a common representation for spoken language understanding . it supports cross-domain queries, fine-grained types, complex utterances and composition . the proposed representation was released to developers at a trade show in 2016 . |
Prompt-free and Efficient Few-shot Learning with Language Models (2022.acl-long)
Copied to clipboard
Rabeeh Karimi Mahabadi, Luke Zettlemoyer, James Henderson, Lambert Mathias, Marzieh Saeidi, Veselin Stoyanov, Majid Yazdani
| Challenge: | Existing methods for few-shot fine-tuning of pretrained language models require carefully engineered prompts and verbalizers to convert inputs into a cloze-format that the PLM can score. |
| Approach: | They propose a method for few-shot fine-tuning of pretrained language models that uses task-specific adapters instead of manually engineered prompts and verbalizers. |
| Outcome: | The proposed method outperforms existing state-of-the-art methods on a wide range of few shot NLP tasks. |
TRACE: A Framework for Analyzing and Enhancing Stepwise Reasoning in Vision-Language Models (2026.eacl-long)
Copied to clipboard
| Challenge: | Evaluating large vision-language models has focused on final-answer correctness, but this metric is often insufficient and misleading. |
| Approach: | They propose a framework that decomposes complex multimodal tasks into Auxiliary Reasoning Sets (ARS) ARS decomposition reveals how consistently a model reasons across sub-questions with structured dependencies. |
| Outcome: | a new framework improves diagnostic evaluation of large vision-language models . it decomposes complex multimodal tasks into auxiliary reasoning sets with structured dependencies . the framework pinpoints reasoning failures and exposes errors overlooked by standard evaluation . |
Meta-training with Demonstration Retrieval for Efficient Few-shot Learning (2023.findings-acl)
Copied to clipboard
| Challenge: | Large language models have impressive fewshot performance on many NLP tasks and domains. |
| Approach: | They propose a meta-training approach that uses demonstration retrieval to train parameter-efficient models that generalize well on a larger variety of tasks. |
| Outcome: | The proposed approach outperforms many parameter-efficient methods on QA, NLI, and text classification tasks. |
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. |
Scaling Multi-Domain Dialogue State Tracking via Query Reformulation (N19-2)
Copied to clipboard
| Challenge: | Using a pointer-generator network, we model the reference resolution task as a dialogue context-aware user query reformulation task. |
| Approach: | They propose a pointer-generator network and a novel multi-task learning setup to model dialogue state tracking and referring expression resolution tasks using a dialogue context-aware user query reformulation task. |
| Outcome: | The proposed model improves absolute F1 on internal and public benchmarks. |
TimelineQA: A Benchmark for Question Answering over Timelines (2023.findings-acl)
Copied to clipboard
Wang-Chiew Tan, Jane Dwivedi-Yu, Yuliang Li, Lambert Mathias, Marzieh Saeidi, Jing Nathan Yan, Alon Halevy
| Challenge: | Existing question answering techniques for lifelogs do not provide accurate answers . augmented reality glasses have led to the creation of personal assistants . |
| Approach: | They propose to use a benchmark to query lifelogs to find out what happened in real life . they find that extractive QA systems out-perform retrieval-augmented QA techniques . |
| Outcome: | The proposed method outperforms state-of-the-art retrieval-augmented QA systems in atomic queries and multi-hop queries. |
ToKen: Task Decomposition and Knowledge Infusion for Few-Shot Hate Speech Detection (2022.emnlp-main)
Copied to clipboard
Badr AlKhamissi, Faisal Ladhak, Srinivasan Iyer, Veselin Stoyanov, Zornitsa Kozareva, Xian Li, Pascale Fung, Lambert Mathias, Asli Celikyilmaz, Mona Diab
| Challenge: | Hate speech detection is complex and requires commonsense reasoning and social nuance . prior work has shown that even humans cannot achieve a high agreement on whether a post constitutes HS . |
| Approach: | They frame a few-shot learning task to decompose a hate speech detection task into its "constituent" parts. they show that infusing commonsense knowledge from reasoning datasets improves the performance even further. |
| Outcome: | The proposed method outperforms baseline methods in the 16-shot case. |