mlsanityData to 7 Oct 2026

ReContext: Recursive Evidence Replay as LLM Harness for Long-Context Reasoning

arXiv 2607.02509 ↗First posted 2 Jul 2026cs.AIcited by 2

On the map: LLM reasoning → Test-time compute and efficient thinking

Abstract (arXiv)

Understanding and reasoning over long contexts has become a key requirement for deploying large language models (LLMs) in realistic applications. Although recent LLMs support increasingly long context windows, they often fail to use relevant evidence that is already present in the input, revealing a gap between context access and effective context utilization. In this work, we propose Recursive Evidence Replay as LLM Harness for Long-Context Reasoning (RECONTEXT), a training-free inference method for improving long-context reasoning. RECONTEXT uses model-internal relevance signals to construct a query-conditioned evidence pool and replays it before final generation while preserving the full original context. This recursive selection process separates evidence organization from answer generation without training, external memory, or context pruning. We also provide a theoretical analysis based on associative memory, which characterizes the context as a memory store, the question as a retrieval cue, attention as cue-trace association, and replay as trace reactivation. Experiments on eight long-context datasets with 128K context length show that RECONTEXT consistently improves evidence utilization across Qwen3-4B, Qwen3-8B, and Llama3-8B, achieving the best average rank on all three backbones. Code is available at https://github.com/Yanjun-Zhao/ReContext.

Key facts. Cited by 2 papers in the map (data through 7 Oct 2026). Placed in Test-time compute and efficient thinking, one of 17 directions in LLM reasoning. Builds on 41 of its 53 references that are in the map. First posted on arXiv 2 Jul 2026.

Builds on

41 of 53 references.

Closest ancestors · 1

Same direction, newest first

  1. 2025SABER: Switchable and Balanced Training for Efficient LLM Reasoning8 Aug

Other references · 40

  1. 2020Retrieval-Augmented Generation for Knowledge-Intensive NLP Taskscited by 9,160
  2. 2023Lost in the Middle: How Language Models Use Long Contextscited by 3,383
  3. 2020Longformer: The Long-Document Transformercited by 3,362
  4. 2023Efficient Streaming Language Models with Attention Sinkscited by 2,115
  5. 2020Big Bird: Transformers for Longer Sequencescited by 1,673
  6. 2023LongBench: A Bilingual, Multitask Benchmark for Long Context Understandingcited by 1,378
  7. 2024RULER: What's the Real Context Size of Your Long-Context Language Models?cited by 1,270
  8. 2025A-MEM: Agentic Memory for LLM Agentscited by 1,000
  9. 2022In-context Learning and Induction Headscited by 930
  10. 2024SnapKV: LLM Knows What You are Looking for Before Generationcited by 848
  11. 2023H2O: Heavy-Hitter Oracle for Efficient Generative Inference of Large Language Modelscited by 805
  12. 2023YaRN: Efficient Context Window Extension of Large Language Modelscited by 584
28 more
  1. 2024Quest: Query-Aware Sparsity for Efficient Long-Context LLM Inferencecited by 487
  2. 2020Hopfield Networks is All You Needcited by 461
  3. 2023LongLLMLingua: Accelerating and Enhancing LLMs in Long Context Scenarios via Prompt Compressioncited by 443
  4. 2023Scissorhands: Exploiting the Persistence of Importance Hypothesis for LLM KV Cache Compression at Test Timecited by 428
  5. 2024PyramidKV: Dynamic KV Cache Compression based on Pyramidal Information Funnelingcited by 420
  6. 2024∞Bench: Extending Long Context Evaluation Beyond 100K Tokenscited by 360
  7. 2014Memory Networkscited by 354
  8. 2023LLMLingua: Compressing Prompts for Accelerated Inference of Large Language Modelscited by 279
  9. 2023LongLoRA: Efficient Fine-tuning of Long-Context Large Language Modelscited by 243
  10. 2023L-Eval: Instituting Standardized Evaluation for Long Context Language Modelscited by 232
  11. 2024Ada-KV: Optimizing KV Cache Eviction by Adaptive Budget Allocation for Efficient LLM Inferencecited by 230
  12. 2023LongNet: Scaling Transformers to 1,000,000,000 Tokenscited by 201
  13. 2023Compressing Context to Enhance Inference Efficiency of Large Language Modelscited by 185
  14. 2023ZeroSCROLLS: A Zero-Shot Benchmark for Long Text Understandingcited by 175
  15. 2024HELMET: How to Evaluate Long-Context Language Models Effectively and Thoroughlycited by 119
  16. 2024Second-Order Fine-Tuning without Pain for LLMs:A Hessian Informed Zeroth-Order Optimizercited by 51
  17. 2026Mem-Gallery: Benchmarking Multimodal Long-Term Conversational Memory for MLLM Agentscited by 49
  18. 2026A Survey of Agentic Reasoning for Large Language Models: Towards Recursively Self-Improving and Collective Agentscited by 47
  19. 2026Code as Agent Harnesscited by 40
  20. 2025Modern Methods in Associative Memorycited by 12
  21. 2025RiskPO: Risk-based Policy Optimization via Verifiable Reward for LLM Post-Trainingcited by 12
  22. 2025FZOO: Fast Zeroth-Order Optimizer for Fine-Tuning Large Language Models towards Adam-Scale Speedcited by 11
  23. 2025CLIPPER: Compression enables long-context synthetic data generationcited by 9
  24. 2024Sparse-VQ Transformer: An FFN-Free Framework with Vector Quantization for Enhanced Time Series Forecastingcited by 6
  25. 2023Does Vector Quantization Fail in Spatio-Temporal Forecasting? Exploring a Differentiable Sparse Soft-Vector Quantization Approachcited by 3
  26. 2026PaperMind: Benchmarking Agentic Reasoning and Critique over Scientific Papers in Multimodal LLMscited by 3
  27. 2026DySCO: Dynamic Attention-Scaling Decoding for Long-Context Language Modelscited by 2
  28. 2024Less is more: Embracing sparsity and interpolation with Esiformer for time series forecastingcited by 1

Built on by

Cited by year (2026 to 7 Oct)

Citing papers per year: 2026: 2.

2 papers.

  1. 2026Trace as State: Reasoning Traces as Conditional States for Long-Context Transformerscited by 1
  2. 2026A Table Is Worth 64 Tokens: Pixel-level Compression for Multi-Table Document Question Answeringcited by 0

Similar, not cited

Close in wording, no citation either way

  1. 2026RECAST: Learning to Compute the Right Context through Adaptive Evidence Routing
  2. 2025Improving Context Fidelity via Native Retrieval-Augmented Reasoning
  3. 2026RecaLLM: Addressing the Lost-in-Thought Phenomenon with Explicit In-Context Retrieval
  4. 2024ALR2: A Retrieve-then-Reason Framework for Long-context Question Answering
  5. 2026Chained Recursive Language Models for Multi-Iteration Reasoning
  6. 2025Look Back to Reason Forward: Revisitable Memory for Long-Context LLM Agents
  7. 2026CoEM: Empowering Long-Context Reasoning with Commit-on-Evidence Memory
  8. 2024BABILong: Testing the Limits of LLMs with Long Context Reasoning-in-a-Haystack