Reinforcement Learning for Self-Guided Context Compression in Mathematical Reasoning

Better Starts, Better Ends: Bootstrapped Iterative Self-Reasoning Distillation for Compressed Reasoning

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.

Structured Thoughts For Improved Reasoning And Context Pruning

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.

Leveraging Instruction Tuning and Merging for Reasoning Model Adaptation

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.

MathCritique: Enhancing LLM Reasoning via Critique Models with Test-Time and Training-Time Supervision

ReToolSQL: Agentic Reinforcement Learning for Robust Text-to-SQL

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 practical path toward robust enterprise-grade text-to-SQL.

AdaptiReason: Adaptive Chain-of-Thought Compression for Efficient Large Language Model Inference

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-constrained environments. This paper proposes AdaptiReason, a novel framework that dynamically compresses intermediate reasoning steps based on task difficulty and model confidence without requiring retraining of the underlying base model. AdaptiReason employs a lightweight difficulty estimator to determine the appropriate reasoning depth for each input, followed by a learned token-pruning policy that eliminates redundant or low-information reasoning steps. Experimental evaluation on the MATH, GSM8K, and ARC-Challenge benchmarks demonstrates that AdaptiReason reduces the average number of generated tokens by 3.7× while preserving 98.2% of the baseline reasoning accuracy. Furthermore, the proposed framework is model-agnostic and can be seamlessly integrated with instruction-tuned LLMs without requiring access to model parameters, relying solely on output logits for adaptive reasoning compression. The results demonstrate that AdaptiReason significantly improves inference efficiency while maintaining high reasoning performance, making it suitable for real-time and resource-constrained LLM applications.