AI Paper Atlas

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

SoLoPO: Unlocking Long-Context Capabilities in LLMs via Short-to-Long Preference Optimization

Huashan Sun, Shengyi Liao, Yansen Han, Yu Bai, Yang Gao, Cheng Fu, Weizhou Shen, Fanqi Wan, Ming Yan, Ji Zhang, Fei Huang

ICLR 2026 · Large language models › Language, safety and interpretability · approach: LLM pruning and scaling data · 8 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
replace long-context ORPO with short-to-long SoLo-ORPOORPO on Qwen2.5-7B · LongBenchV1 QA avgF1: 45.3 -> 49.5Improved
replace long-context ORPO with SoLo-ORPOORPO on Qwen2.5-7B · RULER QA avgaccuracy: 55.1 -> 65.1Improved
replace long-context DPO with SoLo-DPODPO on Qwen2.5-7B · LongBenchV1 QA avgF1: 46.9 -> 47.8Improved
replace long-context DPO with SoLo-DPODPO on Qwen2.5-7B · RULER QA avgaccuracy: 62.2 -> 62.8Improved
replace long-context SimPO with SoLo-SimPOSimPO on Qwen2.5-7B · LongBenchV1 QA avgF1: 44.6 -> 48.4Improved
replace long-context SimPO with SoLo-SimPOSimPO on Qwen2.5-7B · RULER QA avgaccuracy: 57.5 -> 63.9Improved
replace long-context ORPO with SoLo-ORPOORPO on Llama3.1-8B · LongBenchV1 QA avgF1: 45.4 -> 46.3Improved
replace long-context ORPO with SoLo-ORPOORPO on Llama3.1-8B · RULER QA avgaccuracy: 58.9 -> 61.3Improved
replace long-context DPO with SoLo-DPODPO on Qwen2.5-7B · LongBenchV2 overallaccuracy: 29.7 -> 35.2Improved
replace long-context ORPO with SoLo-ORPOORPO on Qwen2.5-7B · LongBenchV2 overallaccuracy: 26.6 -> 33.2Improved
replace long-context SimPO with SoLo-SimPOSimPO on Qwen2.5-7B · LongBenchV2 overallaccuracy: 25.4 -> 31.0Improved
apply SoLo-DPO versus untuned Instruct modelQwen2.5-7B-Instruct · Open LLM Leaderboard avgaverage score: 48.56 -> 48.12Hurt

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