12 papers · 1 filter
A Safe Action Is Not Enough: Feasible-Future Decoding for Vision-Language-Action Policies
Tu Nguyen, Matthieu Zimmer, Vu Anh Vu +4
A safe action is not necessarily a viable one. A frozen vision-language-action (VLA) policy can favor a locally admissible move that leaves no policy-supported route to safe task c…
The Weakest Link: Distilling LLM Reasoning with Worst-Case Constrained Reinforcement Learning
Matthieu Zimmer, Xiaotong Ji, Tu Nguyen +1
Distilling the reasoning capabilities of large language models (LLMs) into smaller students is a central challenge for efficient deployment. Current approaches face a fundamental t…
Tail-Influence Sampling for CVaR Policy Evaluation
Pauline Bourigault, Xiaotong Ji, Matthieu Zimmer +2
Policies with similar mean returns can differ sharply in rare failures, yet estimating lower-tail conditional value-at-risk (CVaR) accurately can require many costly rollouts. When…
Beam Search as Test-Time Self-Distillation via Counterfactual Contexts
Su Ee Tan, Xiaotong Ji, Rasul Tutunov +2
Self-Distillation Fine-Tuning (SDFT) enables a language model to act as its own teacher: by conditioning on a demonstration, the model produces an implicit reward via pointwise mut…
Composable Decoding on the Probability Simplex: Theory and Implementation
Xiaotong Ji, Ahmed Khaled Khamis, Rasul Tutunov +2
Decoding for large language models is typically treated as a collection of isolated sampling strategies, with limited theoretical understanding of the behaviours they induce and ho…
iSDFT: Information-Proximal Self-Distillation for Continual Learning in LLMs
Ahmed Khaled Khamis, Xiaotong Ji, Hassan Jaber +4
On-policy self-distillation fine-tuning (SDFT) learns new skills from demonstrations while reducing forgetting, but it always distils toward the full demonstration-conditioned teac…
The Model Knows, the Decoder Finds: Future Value Guided Particle Power Sampling
Tu Nguyen, Matthieu Zimmer, Rasul Tutunov +2
A recurring pattern in "reasoning without training" is that base LLMs already assign non-trivial probability mass to correct multi-step solutions; the bottleneck is locating these…
Risk-Controlled Lean-as-Judge for Natural-Language Mathematical Reasoning
Pauline Bourigault, Xiaotong Ji, Matthieu Zimmer +2
Lean is increasingly used to judge natural-language mathematical answers, but its signal is partial: many answers never formalize, and a failed proof may reflect an ill-typed state…
The -Combinator for LLMs: Solving Long-Context Rot with -Calculus
Amartya Roy, Rasul Tutunov, Xiaotong Ji +2
LLMs are increasingly used as general-purpose reasoners, but long inputs remain bottlenecked by a fixed context window. Recursive Language Models (RLMs) address this by externalisi…
Decoding as Optimisation on the Probability Simplex: From Top-K to Top-P (Nucleus) to Best-of-K Samplers
Xiaotong Ji, Rasul Tutunov, Matthieu Zimmer +1
Decoding sits between a language model and everything we do with it, yet it is still treated as a heuristic knob-tuning exercise. We argue decoding should be understood as a princi…
Multi-Task GRPO: Reliable LLM Reasoning Across Tasks
Shyam Sundhar Ramesh, Xiaotong Ji, Matthieu Zimmer +5
RL-based post-training with GRPO is widely used to improve large language models on individual reasoning tasks. However, real-world deployment requires reliable performance across…
Scalable Power Sampling: Unlocking Efficient, Training-Free Reasoning for LLMs via Distribution Sharpening
Xiaotong Ji, Rasul Tutunov, Matthieu Zimmer +1
Reinforcement learning (RL) post-training is a dominant approach for improving the reasoning performance of large language models (LLMs), yet growing evidence suggests that its gai…