Antoine Bosselut

PERK: Long-Context Reasoning as Parameter-Efficient Test-Time Learning

Revisiting Multilingual Data Mixtures in Language Model Pretraining

Mixture of Cognitive Reasoners: Modular Reasoning with Brain-Like Specialization

AbstRaL: Augmenting LLMs' Reasoning by Reinforcing Abstract Thinking

Diversity Matters: Revisiting Test-Time Compute in Vision-Language Models

GaLLoP: Gradient-based Sparse Learning on Low-Magnitude Parameters

Can Performant LLMs Be Ethical? Quantifying the Impact of Web Crawling Opt-Outs

Positional Fragility in LLMs: How Offset Effects Reshape Our Understanding of Memorization Risks

INCLUDE: Evaluating Multilingual Language Understanding with Regional Knowledge

For Better or for Worse, Transformers Seek Patterns for Memorization

LLMs Are In-Context Bandit Reinforcement Learners

LLMs Are In-Context Reinforcement Learners