The paper proposes a new method for training large language models to handle long-context reasoning by combining Group Relative Policy Optimization (GRPO) with on‑policy distillation (OPD). It introduces a synthetic multilingual dataset called LongBlocks that tests multi‑hop reasoning, contextual grounding, and long‑form generation. Experiments show that the combined approach outperforms either GRPO or OPD alone while maintaining short‑context performance.
arXiv:2606. 18831v1 Announce Type: cross Abstract: Long-context reasoning is an essential capability for large language models, particularly when they are deployed as autonomous agents that must reason over lengthy trajectories.
arXiv:2607. 06160v1 Announce Type: cross Abstract: Synthesizing long-context supervised fine-tuning (SFT) data is a scalable way to enhance the long-context understanding of large language models (LLMs), yet existing approaches share three limitations: narrow task coverage, insufficient instruction difficulty, and a lack of faithfulness supervision.
The paper introduces Echo-GRPO, a method that rewrites privileged reasoning traces into a model’s own idiolect to align off‑policy supervision with the student policy’s vocabulary. By preserving semantics through Dual‑Reference Decoding, Echo‑GRPO mitigates gradient clipping on critical reasoning tokens and improves reasoning distillation. The approach is instantiated as VideoEcho‑R1 for video reasoning, yielding consistent gains across multiple multimodal LLM backbones and benchmarks, and it can be applied as a plug‑in to both RL and supervised fine‑tuning frameworks.
arXiv:2606. 07000v1 Announce Type: new Abstract: Recent post-training methods, particularly Reinforcement Learning with Verifiable Rewards (RLVR), have significantly enhanced the reasoning ability of Large Vision-Language Models (LVLMs).
arXiv:2607. 19345v1 Announce Type: cross Abstract: Large language models that generate step-by-step reasoning traces have achieved strong performance on complex tasks, and extending them to long-context settings has emerged as an important frontier.
arXiv:2508. 10123v3 Announce Type: replace-cross Abstract: Advanced reasoning in LLMs on challenging domains like mathematical reasoning can be tackled using verifiable rewards based reinforced fine-tuning (ReFT).
The paper reports that in on‑policy distillation for large language models, reasoning performance can be improved by supervising only a tiny fraction of generated tokens—sometimes just one or two tokens per reasoning trajectory, about 0.05% of all tokens. This sparse supervision consistently matches or exceeds full‑token training across nine teacher‑student setups on mathematical reasoning, and is also validated on coding reasoning, Llama models, and PPO‑based reinforcement learning with verifiable reward. The findings suggest that effective post‑training does not require token‑intensive supervision and may align more closely with natural learning processes that focus on critical reasoning steps.
arXiv:2605. 03677v2 Announce Type: replace Abstract: On-policy distillation (OPD) has recently emerged as an effective post-training paradigm for consolidating the capabilities of specialized expert models into a single student model.
arXiv:2607. 16205v1 Announce Type: new Abstract: Reinforcement learning with verifiable rewards has emerged as a standard approach for enhancing reasoning in large language models, which typically optimizes the policy by contrasting multiple self generated rollouts.
Randomized YaRN is a training method that enhances length generalization for large language models by combining YaRN-based positional extrapolation with randomized positional encoding and a length curriculum. During training on short-context data, tokens receive YaRN positional encodings sampled from a larger position range, exposing the model to out-of-distribution positional representations. Evaluated on BABILong, Multi-Round Coreference Resolution, and LongBench v2, Randomized YaRN consistently improves reasoning performance on context lengths from 16K to 128K, outperforming standard fine‑tuning especially at far out‑of‑distribution lengths.
ContextPilot is a proactive context‑management framework designed to improve long‑horizon agentic reasoning with large language models. It expands the toolset to include planning, long‑term memory, and soft context offloading, and introduces a reinforcement‑learning strategy that focuses on critical editing decisions and assigns action‑level advantages. Experiments on long‑context QA and deep search tasks demonstrate that ContextPilot achieves stronger performance with a more compact working context, outperforming existing baselines across various base models and benchmarks.
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.