TL;DR
This project tackles the Countdown math reasoning problem, and aims to train a model to use a tool which summarizes its own reasoning context, thereby compacting its reasoning trace.
BIRD(Bootstrapped Iterative Self-Reasoning Distillation), a two-stage self-reasoning distillation method that improves the rollout distribution before on-policy training and achieves a stronger accuracy-efficiency trade-off than prompting and cold-start on-policy distillation on MATH-500 and AIME benchmarks.
This work introduces Structured Thoughts, a framework that organizes reasoning into alternating blocks and blocks that captures exploratory scratch work, while the distilled conclusion of that step contains the distilled conclusion of that step.
This work uses classic instruction tuning, supervised fine-tuning without reasoning traces, on the RLM to improve RLM performance in both verifiable and hard-to-verify domains, including coding and text summarization, while preserving RLM capabilities across other domains.
A critique-in-the-loop self-improvement method that incorporates critique-based supervision into the actor’s self-training process and improves the actor’s exploration efficiency and solution diversity, especially on challenging queries, leading to a stronger actor model.
ReToolSQL is presented, a two-stage training framework for text-to-SQL that combines a supervised warm-start on rejection-sampled reasoning traces with agentic reinforcement fine-tuning (RFT) over multi-turn tool-use trajectories and shows that a properly designed SFT$\to-RFT pipeline over tool-use trajectories is a pr...
Large Language Models (LLMs) trained using Chain-of-Thought (CoT) supervision have achieved state-of-the-art performance on complex reasoning tasks. However, the generation of long reasoning chains introduces substantial computational overhead during inference, limiting their deployment in low-latency and resource-cons...
We use cookies to run the site and, with your consent, for analytics and to show ads. See our Cookie Policy.
