AI Paper Atlas

Research map / Large language models / Language, safety and interpretability

EntropyLong: Effective Long-Context Training via Predictive Uncertainty

Junlong Jia, Ziyang Chen, Xing W, Chaochen Gao, Zijia Lin, Songlin Hu, GuoBinghui

ICLR 2026 · Large language models › Language, safety and interpretability · approach: LLM pruning and scaling data · 6 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
entropy-verified long-range dependency data (EntropyLong) vs NExtLongLlama-3-8B · RULER avgscore: 85.22 -> 87.37Improved
EntropyLong vs NExtLong at 128kLlama-3-8B · RULER 128kscore: 77.99 -> 81.26Improved
EntropyLong data then SFTLlama-3-8B · LongBench-v2 overallaccuracy: 24.10 -> 27.60Improved
EntropyLong data then SFTLlama-3-8B · LongBench-v2 Longaccuracy: 23.10 -> 31.50Improved
add empirical entropy-reduction verification (ablation)EntropyLong · RULER avgscore: 85.82 -> 87.37Improved
use alpha=2.0 instead of 1.5 (ablation)EntropyLong · RULER avgscore: 82.49 -> 87.37Improved
use alpha=2.0 instead of 2.5 (ablation)EntropyLong · RULER avgscore: 85.52 -> 87.37Improved
use verification threshold 0.4 instead of 0.2 (ablation)EntropyLong · RULER avgscore: 85.45 -> 87.37Improved
use verification threshold 0.4 instead of 0.6 or 0.8 (ablation)EntropyLong · RULER avgscore: 86.14 / 86.47 -> 87.37Improved
window size 16 instead of 2, 4, 8 (ablation)EntropyLong · RULER avgscore: 86.04-86.85 -> 87.37Mixed
random shuffle instead of sequence concatenation (ablation)EntropyLong · RULER avgscore: 87.06 -> 87.37No clear effect
train with EntropyLong dataLlama-3-8B · short-text benchmarks avgaccuracy: 61.77 -> 61.83No clear effect

Earlier methods this paper compares with

Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.

Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy