mlsanityData to 14 Oct 2026

PolicyLong: Towards On-Policy Context Extension

arXiv 2604.07809 ↗First posted 9 Apr 2026cs.LG cs.AIcited by 0

Not on the map. Why?

Abstract (arXiv)

Extending LLM context windows is hindered by scarce high-quality long-context data. Recent methods synthesize data with genuine long-range dependencies via information-theoretic verification, selecting contexts that reduce a base model's predictive entropy. However, their single-pass offline construction with a fixed model creates a fundamental off-policy gap: the static screening landscape misaligns with the model's evolving capabilities, causing the training distribution to drift. We propose PolicyLong, shifting data construction towards a dynamic on-policy paradigm. By iteratively re-executing data screening (entropy computation, retrieval, and verification) using the current model, PolicyLong ensures the training distribution tracks evolving capabilities, yielding an emergent self-curriculum. Crucially, both positive and hard negative contexts derive from the current model's entropy landscape, co-evolving what the model learns to exploit and resist. Experiments on RULER, HELMET, and LongBench-v2 (Qwen2.5-3B) show PolicyLong consistently outperforms EntropyLong and NExtLong, with gains growing at longer contexts (e.g., +2.54 at 128K on RULER), confirming the value of on-policy data evolution.

Key facts. No citing papers in the map yet (data through 14 Oct 2026). Builds on 24 of its 29 references that are in the map. First posted on arXiv 9 Apr 2026.

Builds on

24 of 29 references.

References · 24

  1. 2020A Simple Framework for Contrastive Learning of Visual Representationscited by 12,049
  2. 2021RoFormer: Enhanced Transformer with Rotary Position Embeddingcited by 4,703
  3. 2020Dense Passage Retrieval for Open-Domain Question Answeringcited by 3,824
  4. 2022FlashAttention: Fast and Memory-Efficient Exact Attention with IO-Awarenesscited by 3,673
  5. 2023Lost in the Middle: How Language Models Use Long Contextscited by 3,455
  6. 2017Billion-scale similarity search with GPUscited by 2,497
  7. 2024RULER: What's the Real Context Size of Your Long-Context Language Models?cited by 1,346
  8. 2020Approximate Nearest Neighbor Negative Contrastive Learning for Dense Text Retrievalcited by 849
  9. 2020Contrastive Learning with Hard Negative Samplescited by 628
  10. 2020RocketQA: An Optimized Training Approach to Dense Passage Retrieval for Open-Domain Question Answeringcited by 413
  11. 2024LongBench v2: Towards Deeper Understanding and Reasoning on Realistic Long-context Multitaskscited by 379
  12. 2026Reinforcement Learning via Self-Distillationcited by 332
12 more
  1. 2023Effective Long-Context Scaling of Foundation Modelscited by 303
  2. 2026Self-Distillation Enables Continual Learningcited by 247
  3. 2024jina-embeddings-v3: Multilingual Embeddings With Task LoRAcited by 141
  4. 2024HELMET: How to Evaluate Long-Context Language Models Effectively and Thoroughlycited by 125
  5. 2023In-context Pretraining: Language Modeling Beyond Document Boundariescited by 79
  6. 2024What is Wrong with Perplexity for Long-context Language Modeling?cited by 69
  7. 2026POPE: Learning to Reason on Hard Problems via Privileged On-Policy Explorationcited by 42
  8. 2025NExtLong: Toward Effective Long-Context Training without Long Documentscited by 20
  9. 2024Quest: Query-centric Data Synthesis Approach for Long-context Scaling of Large Language Modelcited by 16
  10. 2025LongMagpie: A Self-synthesis Method for Generating Large-scale Long-context Instructionscited by 6
  11. 2025EntropyLong: Effective Long-Context Training via Predictive Uncertaintycited by 5
  12. 2025LiteLong: Resource-Efficient Long-Context Data Synthesis for LLMscited by 3

Built on by

Not cited yet.