ReContext: Training-Free Long-Context Inference
burkov · x · 2026-07-13
This paper/method introduces ReContext, a training-free long-context inference method.
The core idea is to use the model's internal correlation signals for recursive evidence replay, repeatedly feeding key evidence back to the model to enhance reasoning in long texts. The authors claim it achieves SOTA across multiple models and datasets and scales to a 128K context length.
It's highly relevant for readers interested in long-context handling, reasoning enhancement, and inference-stage optimization.
More from Research
- Cognition's SWE-2 uses a KKT duality argument in RL to shift the effort Pareto curve — YouJiacheng · 2026-09-11
- VidMap uses RoMa coarse matching on all frames, fine-scale only for keyframes — ducha_aiki · 2026-09-11
- Bug Hunt Bench author: leaderboard noise is about 2-3 points — PawelHuryn · 2026-09-11
- PNAS paper shows a tiny billiard-ball system is a universal computer — undecidability lives in two dimensions — eigensteve · 2026-09-11
- New paper: Absolute pose estimation from affine cues and gravity direction — ducha_aiki · 2026-09-11
- LoMa Paper Ships REALLY HardPairs Dataset, Accepted at ECCV 2026 — ducha_aiki · 2026-09-11