AI Paper Atlas

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

LLMInertia: Adaptive Counter-Inertial Reasoning to Improve Evidence Faithfulness in Large Language Models

Xinxin You, Xien Liu, Chenwei Yan, Siqi Song, Chen Ning, Kaiyin Zhou, Shaohui Liu, Ji Wu

ICML 2026 · Large language models › Language, safety and interpretability · approach: Faithfulness of LLM explanations

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 adaptive counter-inertial reminders (LLMInertia) to the base modelLLaMA-3-Instruct-8B · CNN/DailyMailUniEval Consistency: 86.40 -> 90.33Improved
add LLMInertia to the base modelQwen-2.5-Instruct-7B · SQuAD v2Anah-v2: 17.22 -> 11.22Improved
add LLMInertia to the base modelLLaMA-3-Instruct-70B · SQuAD v2Anah-v2: 11.52 -> 9.56Improved
add LLMInertia instead of SFTLLaMA-3-Instruct-8B · HaluEvalAnah-v2: 8.27 vs 8.21No clear effect
add entity extraction (vs generic reminder for all samples) (ablation)LLMInertia · CNN/DailyMailUniEval Consistency: 87.54 -> 90.33Improved
add the LLM inertia probe (vs sliding-window co-occurrence pairs) (ablation)LLMInertia · CNN/DailyMailUniEval Consistency: 85.76 -> 90.33Improved
add the LLM inertia probe (vs sliding-window co-occurrence pairs) (ablation)LLMInertia · SQuAD v2Anah-v2: 17.48 -> 12.53Improved
add adaptive reminders (vs fixed generic reminder) (ablation)LLMInertia · SQuAD v2Anah-v2: 14.99 -> 12.53Improved
increase probe output from 1 to 5 co-occurring entities per entity (ablation)LLMInertia · CNN/DailyMailAlignScore: 86.73% -> 88.24%Improved
increase probe output from 3 to 5 entities (ablation)LLMInertia · Medical induction (25%)accuracy: peaks at 3, declines slightly at 5Hurt
increase proportion of co-occurrence data in continued pretrainingLLaMA-3-8B (medical) · Contradiction testhallucination induction rate: 15.71% -> 44.56%Hurt
increase proportion of co-occurrence data in continued pretrainingLLaMA-3-8B (medical) · Contradiction testaccuracy: decreases by 9.38% to 37.70%Hurt

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