AI Paper Atlas

Research map / Large language models / Instruction and fine-tuning

SEMA: Simple yet Effective Learning for Multi-Turn Jailbreak Attacks

Mingqian Feng, Xiaodong Liu, Weiwei Yang, Jialin Song, Xuekai Zhu, Chenliang Xu, Jianfeng Gao

ICLR 2026 · Large language models › Instruction and fine-tuning · approach: Safety alignment after fine-tuning · 10 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
add intent alignment score to the reward (ablation)SEMA · AdvBench (LLM Classifier, mean over 3 victims)ASR@1 (%): 50.4 -> 67.5Improved
use No-Refusal Indicator as the reward instead of intent-aligned reward (ablation)SEMA · AdvBench (LLM Classifier)ASR@1 (%): zero success rateHurt
add prefilling self-tuning stage (ablation)SEMA with Llama-3.2-3B-Instruct attacker · training rewardreward curveImproved
increase training max turns from 1 up to 7 (ablation)SEMA · AdvBench/HarmBench vs Qwen2.5-3BASR@1 (%)Improved
increase training max turns from 7 to 10 (ablation)SEMA · AdvBench/HarmBench vs Qwen2.5-3BASR@1 (%)Hurt
use SFT instead of DPO offline variantsoffline baselines · AdvBenchASR@1 (%): 32.3/21.0/16.5 -> 38.5/30.6/23.8Improved
use online RL (SEMA) instead of SFT on successful rolloutsSFT baseline · AdvBench (LLM Classifier, mean)ASR@1 (%): 31.0 -> 80.1Improved
use SEMA instead of ActorAttack under non-No-Refusal judgesActorAttack · AdvBench on GPT-oss-20B (HarmBench Classifier)ASR@1 (%): 19.2 -> 57.7Improved
increase attempts from N=5 to N=20SEMA · HarmBench vs GPT-4.1-miniASR@N (%): 96.8 -> 99.7Improved

Later papers that compared with this method

1 later accepted paper reports SEMA as a baseline in their results tables; 1 of them report a better number on most of the results they share with it.

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