AI Paper Atlas

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

Old Habits Die Hard: How Conversational History Geometrically Traps LLMs

Adi Simhi, Fazl Barez, Martin Tutek, Yonatan Belinkov, Shay B Cohen

ICML 2026 · Large language models › Language, safety and interpretability · approach: Bias measurement and mitigation · 0 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
shuffle questions randomly (topic-inconsistent) instead of ordering by semantic similarity (ablation)LLM multi-turn conversations · six datasets, open modelsTr(T) carryover: Tr(T) NaturalQA 1.13 -> 1.00; Sorry 1.57 -> 1.09Hurt
topic-inconsistent conversations vs consistent (ablation)LLM multi-turn conversations · Sorry (open models)P(refusal | refusal): 0.90 -> 0.75Hurt
topic-inconsistent conversations vs consistent (ablation)LLM multi-turn conversations · theta_ref, LLaMA-3.1-8B Sorrytheta_ref (geometric separation): 66.52 -> 69.04No clear effect
use hidden states at 85% depth instead of 30%, 50% or 100% (ablation)geometric perspective · probabilistic-geometric correlationSpearman correlation: 0.60Improved
use refusal datasets instead of hallucination datasetsHistory-Echoes · open modelsTr(T): 1.13 -> 1.57Improved
evaluate closed model GPT-5 instead of open modelsprobabilistic perspective · S-posTr(T): 1.33 -> 1.23No clear effect
condition on previous state k=2,3 steps back instead of only 1 (higher-order Markov)probabilistic perspective · six datasetsDelta_kMixed
fix the question and vary preceding state (200-question subset, LLaMA-3.1)probabilistic perspective · NaturalQA and Do-not-answerP(same state | prior state): 74.36 vs 29.06 (NaturalQA); 88.39 vs 16.96 (Do-not-answer)Improved

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