AI Paper Atlas

Research map / Large language models / Alignment and preferences

The Realignment Problem: When Right becomes Wrong in LLMs

Aakash Sen Sharma, Debdeep Sanyal, Manodeep Ray, Vivek Srivastava, Shirish Karande, Murari Mandal

ICML 2026 · Large language models › Alignment and preferences · approach: Active and token-level preference learning · 0 citations (OpenAlex)

Read the paper

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 impact-score weighting instead of uniform weights (ablation)TRACE · SynthValueBench (Qwen2.5-7B)target policy agreement (%): 62.8 -> 70.2Improved
add impact-score weighting instead of uniform weights (ablation)TRACE · SynthValueBench (Qwen2.5-7B)GPQA: 28.4 -> 30.1Improved
add impact-score weighting instead of uniform weights (ablation)TRACE · SynthValueBench (Qwen2.5-7B)HellaSwag: 75.1 -> 77.3Improved
add the triage stage (ablation)TRACE · PKU-SafeRLHF (Llama-3.1-8B)target policy agreement (%): 58.1 -> 70.7Improved
add impact weighting (ablation)TRACE · PKU-SafeRLHF (Llama-3.1-8B)attack success rate (%): 32.1 -> 27.3Improved
add impact weighting (ablation)TRACE · PKU-SafeRLHF (Llama-3.1-8B)target policy agreement (%): 62.8 -> 70.7Improved
add KL regularization (ablation)TRACE · PKU-SafeRLHF (Llama-3.1-8B)MMLU: 64.1 -> 70.2Improved
add KL regularization (ablation)TRACE · PKU-SafeRLHF (Llama-3.1-8B)target policy agreement (%): 71.5 -> 70.7Hurt
remove triage stage (ablation)TRACE · PKU-SafeRLHF (Llama-3.1-8B)attack success rate (%): 24.6 -> 27.3Hurt
use TRACE instead of naive oracle DPOLlama-3.1-8B · PKU-SafeRLHF data scale 5k-20ktarget policy agreement (%): 35.9 -> 55.9, 52.2 -> 70.7Improved
flip 20% of oracle triage labels (ablation)TRACE · policy agreementpolicy agreement (%): 70.2 -> 61.5Hurt
flip 10% of oracle triage labels (ablation)TRACE · policy agreementpolicy agreement (%): 70.2 -> 67.4Hurt

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