Research
A paper is listed when one of its authors is affiliated with the lab. Data: OpenAlex.
Steering off Course: Reliability Challenges in Steering Language ModelsAllen Institute for AI · Annual Meeting of the Association for Computational Linguistics (ACL)3 citationsJan 1, 2025Do Language Models Mirror Human Confidence? Exploring Psychological Insights to Address Overconfidence in LLMsAllen Institute for AI · Findings of the Association for Computational Linguistics: ACL (Findings of ACL)3 citationsJan 1, 2025Understanding Reference Policies in Direct Preference OptimizationAllen Institute for AI · Findings of the Association for Computational Linguistics: NAACL3 citationsJan 1, 2025mFollowIR: A Multilingual Benchmark for Instruction Following in RetrievalAllen Institute for AI · Lecture notes in computer science3 citationsJan 1, 2025OMEGA: Can LLMs Reason Outside the Box in Math? Evaluating Exploratory, Compositional, and Transformative GeneralizationAllen Institute for AI · neural information processing systems2 citationsJan 1, 2025TESS 2: A Large-Scale Generalist Diffusion Language ModelAllen Institute for AI · Annual Meeting of the Association for Computational Linguistics (ACL)2 citationsJan 1, 2025Literature-Grounded Novelty Assessment of Scientific IdeasAllen Institute for AI · Workshop on Scholarly Document Processing (SDP)2 citationsJan 1, 2025DrawEduMath: Evaluating Vision Language Models with Expert-Annotated Students’ Hand-Drawn Math ImagesAllen Institute for AI · Conference of the Nations of the Americas Chapter of the Association for Computational Linguistics: Human Language Technologies (NAACL)2 citationsJan 1, 2025SELFGOAL: Your Language Agents Already Know How to Achieve High-level GoalsAllen Institute for AI · Conference of the Nations of the Americas Chapter of the Association for Computational Linguistics: Human Language Technologies (NAACL)2 citationsJan 1, 2025Transformers as TransducersAllen Institute for AI · Transactions of the Association for Computational Linguistics2 citationsJan 1, 2025Simple Distillation for One-Step Diffusion ModelsAllen Institute for AI · neural information processing systems1 citationJan 1, 2025Efficient Decoding Methods for Language Models on Encrypted DataAllen Institute for AI · International Joint Conference on Natural Language Processing and Conference of the Asia-Pacific Chapter of the Association for Computational Linguistics (IJCNLP-AACL)1 citationJan 1, 2025BLEUBERI: BLEU is a surprisingly effective reward for instruction followingAllen Institute for AI · neural information processing systems1 citationJan 1, 2025FlexOLMo: Open Language Models for Flexible Data UseAllen Institute for AI, Alphabet · neural information processing systems1 citationJan 1, 2025Inference-time Alignment in Continuous SpaceAlibaba, Allen Institute for AI · neural information processing systems1 citationJan 1, 2025Language Modeling by Language ModelsAllen Institute for AI · neural information processing systems1 citationJan 1, 2025Broken Tokens? Your Language Model can Secretly Handle Non-Canonical TokenizationsAllen Institute for AI · neural information processing systems1 citationJan 1, 2025Hybrid Preferences: Learning to Route Instances for Human vs. AI FeedbackAllen Institute for AI · Annual Meeting of the Association for Computational Linguistics (ACL)1 citationJan 1, 2025M-RewardBench: Evaluating Reward Models in Multilingual SettingsAllen Institute for AI, Cohere · Annual Meeting of the Association for Computational Linguistics (ACL)1 citationJan 1, 2025SimulBench: Evaluating Language Models with Creative Simulation TasksAllen Institute for AI · Findings of the Association for Computational Linguistics: NAACL1 citationJan 1, 2025ReIFE: Re-evaluating Instruction-Following EvaluationAllen Institute for AI · Conference of the Nations of the Americas Chapter of the Association for Computational Linguistics: Human Language Technologies (NAACL)1 citationJan 1, 2025IGGA: A Dataset of Industrial Guidelines and Policy Statements for Generative AIsAllen Institute for AI · arXiv (Cornell University)1 citationJan 1, 2025SciArena: An Open Evaluation Platform for Non-Verifiable Scientific Literature-Grounded TasksAllen Institute for AI · neural information processing systemsJan 1, 2025AutoDiscovery: Open-ended Scientific Discovery via Bayesian SurpriseAllen Institute for AI, Alphabet · neural information processing systemsJan 1, 2025A Little Depth Goes a Long Way: The Expressive Power of Log-Depth TransformersAllen Institute for AI · neural information processing systemsJan 1, 2025