SMRC is a new method that aligns large language models with student reasoning for mathematical error correction. It treats student reasoning as a multi‑step decision problem and uses Monte Carlo Tree Search to find optimal correction paths, while a breadth‑first search guided by the model generates reward signals that are back‑propagated to supervise intermediate steps. The authors also introduce the MSEB benchmark of 158 high‑school math problems and a dual evaluation protocol focusing on solution accuracy and correct‑step retention, showing that SMRC outperforms existing methods on several datasets.
arXiv:2603.00578v2 Announce Type: replace-cross Abstract: Long chain-of-thought~(CoT) reasoning improves the problem-solving capability of large reasoning models~(LRMs), but often incurs substantial...
arXiv:2603.00889v2 Announce Type: replace-cross Abstract: Large Language Models (LLMs) have exhibited remarkable reasoning capabilities, largely enabled by supervised fine-tuning (SFT)- and reinforce...
CoTranslator is a method that converts reverse chain-of-thought (CoT) responses into verified forward CoT responses, enabling efficient large-scale generation and verification of training data for reasoning distillation. Trained on a small set of paired reverse and rejection-sampled forward responses, it improves student model accuracy on mathematics benchmarks by up to 5.82 percentage points and retains 95.0% of rejection sampling’s accuracy while using only 35.9% of its compute. The translator also transfers to new datasets, teachers, and STEM problems, boosting average STEM accuracy for 4B and 8B students by 4.88 and 4.31 percentage points, respectively.
Post-training of reasoning language models is commonly driven by supervised distillation and reinforcement learning with verifiable rewards. Distillation often relies on chain-of-thought annotations that are expensive to obtain and may themselves be noisy, incomplete, or partially incorrect; even when the final solution is correct, an imperfect rationale can interfere with learning.
arXiv:2606. 01682v1 Announce Type: cross Abstract: Selecting the best response from multiple small-model samples using a stronger scorer is a simple inference-time strategy, but fails when the small model has already committed to incorrect reasoning paths.
arXiv:2604.02341v2 Announce Type: replace-cross Abstract: Mathematical reasoning in large language models has improved substantially with reinforcement learning using verifiable rewards, where final...
arXiv:2609.38409v1 Announce Type: new Abstract: Recent progress in large language model reasoning has been driven by benchmarks and reinforcement learning environments with automatically verifiable r...
Selecting the best response from multiple small-model samples using a stronger scorer is a simple inference-time strategy, but fails when the small model has already committed to incorrect reasoning paths. PRM guided search avoids this by scoring candidate continuations during generation, but requires a reward model trained with step-level labels.
arXiv:2610.02703v1 Announce Type: new Abstract: On-policy distillation (OPD) improves large language model reasoning by training students on their own rollouts with dense token-wise supervision from...
arXiv:2605. 03862v4 Announce Type: replace Abstract: Reinforcement learning with verifiable rewards has become a common way to improve explicit reasoning in large language models, but final-answer correctness alone does not reveal whether the reasoning trace is faithful, reliable, or useful to the model that consumes it.
arXiv:2510. 00492v3 Announce Type: replace Abstract: The reliability of large language models (LLMs) during test-time scaling is often assessed with \emph{external verifiers} or \emph{reward models} that distinguish correct reasoning from flawed logic.
arXiv:2606. 19327v1 Announce Type: new Abstract: Post-training of reasoning language models is commonly driven by supervised distillation and reinforcement learning with verifiable rewards.