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.
TraceForge introduces a two‑stage training method that generates reasoning traces in the style of a target model family, using supervised learning followed by adversarial refinement. The system produces traces that are judged as authentic with scores comparable to real traces, and when applied to teacher answers, it improves performance on coding benchmarks such as HumanEval+ and MBPP+. The approach addresses the limitation of existing models that only return answers without detailed reasoning, enabling more transparent and learnable outputs.
arXiv:2607. 14552v1 Announce Type: cross Abstract: A standard recipe for distilling the reasoning ability of large language models (LLMs) is to sample chains of thought from the model, keep those that reach the correct final answer, and fine-tune on the survivors.
The paper introduces a self‑distillation method that leverages answer‑only data to improve large language models (LLMs) after training. By training an inverse reasoning LLM to generate latent chain‑of‑thought traces conditioned on the answer, the authors can perform offline distillation that significantly boosts performance on hard prompts for tasks like Countdown and GSM8K. Additionally, the inverse model serves as an accurate verifier, enhancing best‑of‑n selection accuracy.
Cliff is a reward‑shaping method for reinforcement learning with verifiable rewards that identifies the first mistake in a language model’s reasoning process using an off‑the‑shelf teacher. By splitting each rollout into a correct prefix and an incorrect suffix, Cliff assigns token‑level advantages—positive for correct tokens and negative for the rest—providing fine‑grained supervision. Across 12 scenarios, Cliff outperforms on‑policy distillation by 15% and standard GRPO by 7%, even when the teacher is only modestly capable.
The paper examines on‑policy self‑distillation, where a teacher scores a student’s rollout to provide dense token‑level credit. Experiments show that the raw scoring signals neither localize errors nor effectively penalize incorrect tokens, because locating errors is a search problem that the teacher is not explicitly asked to perform. Introducing a critique‑conditioned approach—first generating a critique of the rollout and then conditioning the teacher on it—produces signals that accurately localize errors, correct erroneous spans, and guide the policy toward backtracking tokens, improving performance on math problems and in cold‑start GRPO settings.
The paper introduces TRACE, a digital‑advertising diagnostic environment that uses simulated interventions to generate verifiable rewards for training reasoning agents. By injecting controlled interventions into a simulator, the hidden cause of anomalies becomes an oracle label, enabling agents to learn to identify root causes and affected segments through noisy, confounded evidence. Experiments show that reinforcement learning with these synthesized rewards outperforms large prompted baselines, achieving higher accuracy while using fewer tool calls.
Cliff is a reward‑shaping method for reinforcement learning with verifiable rewards that identifies the first mistake in a language model’s reasoning process using an off‑the‑shelf teacher. It splits each rollout into a correct prefix and an incorrect suffix, assigning positive token‑level advantages to the prefix and negative feedback to the suffix. Across 12 scenarios, Cliff improves reasoning performance, outperforming on‑policy distillation by 15% and standard GRPO by 7%, even with modest teachers.
arXiv:2605.11467v2 Announce Type: replace-cross Abstract: Reasoning models post-hoc rationalize answers they have already committed to internally, producing chains of *reasoning theater*: deliberativ...
arXiv:2610.19298v1 Announce Type: new Abstract: Reinforcement learning for code usually rewards passing tests, which a model can do while remaining wrong on the inputs the tests do not cover. A progr...
arXiv:2609.38107v1 Announce Type: cross Abstract: Chain-of-thought traces are widely read as records of how models reach their answers, informing debugging, agent auditing, and claims about reasoning...
arXiv:2610.02700v1 Announce Type: new Abstract: On-policy self-distillation (OPSD) supplies dense token-level feedback on trajectories sampled from the student's own policy, a richer training signal...