AI Paper Atlas

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

Activation Steering for LLM Alignment via a Unified ODE-Based Framework

Hongjue Zhao, Haosen Sun, Jiangtao Kong, Xiaochang Li, Qineng Wang, Liwei Jiang, Qi Zhu, Tarek F. Abdelzaher, Yejin Choi, Manling Li, Huajie Shao

ICLR 2026 · Large language models › Language, safety and interpretability · approach: Task vectors and steering

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
use multi-step ODE solving instead of one step (ablation)ODESteer (Falcon-7B) · TruthfulQAT x I: 40.8 -> 42.2Improved
use multi-step ODE solving instead of one step (ablation)ODESteer (Mistral-7B) · TruthfulQAT x I: 58.1 -> 59.9Improved
use multi-step ODE solving instead of one step (ablation)ODESteer (LLaMA3.1-8B) · TruthfulQAT x I: 62.1 -> 63.2Improved
use multi-step ODE solving instead of one step (ablation)ODESteer (Falcon-7B) · UltraFeedbackwin rate: 54.0 -> 56.3Improved
use multi-step ODE solving instead of one step (ablation)ODESteer (Falcon-7B) · RealToxicityPromptstoxicity (lower better): 0.199 -> 0.188Improved
use nonlinear density-ratio features (one-step) instead of linear ITI (ablation)One-step ODESteer (Mistral-7B) · TruthfulQAT x I: 46.4 -> 58.1Improved
use nonlinear features (one-step) instead of linear ITI (ablation)One-step ODESteer (Mistral-7B) · RealToxicityPromptstoxicity (lower better): 0.165 -> 0.113Improved
use nonlinear features (one-step) instead of linear ITI (ablation)One-step ODESteer (Falcon-7B) · UltraFeedbackwin rate: 50.5 -> 54.0Improved
use RK4 solver instead of Euler (ablation)ODESteer (LLaMA3.1-8B) · TruthfulQAT x I: 63.2 -> 63.3No clear effect
use RK4 solver instead of Euler (ablation)ODESteer (Falcon-7B) · TruthfulQAT x I: 42.2 -> 42.8No clear effect
apply ODESteer trained on TruthfulQA to MMLUODESteer (LLaMA3.1-8B) · MMLUaccuracy: 61.8 -> 60.9No clear effect
apply ODESteer trained on TruthfulQA to CommonsenseQAODESteer (LLaMA3.1-8B) · CommonsenseQAaccuracy: 68.0 -> 68.3No 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