Causal Interventions in Language Models

Follow this topic Ask about this topic

Momentum

55 papers in the last four weeks, up 162% on the four weeks before. 0.5% of all new papers.

Jul 13Week of Sep 28

Latest papers 317

All topics

Right Knowledge, Wrong Answer: Test-Time Steering for Temporal Fact Conflicts in Open-Weight Language Models
Jun 18, 2026Elias Hossain, Sourav Saha, Umesh Chandra Biswas +1Knowledge Conflicts in Language Models Temporal Reasoning in Language Models
University of Central Florida · 2Mississippi State University
Actionable Activation Directions for Detecting and Mitigating Emergent Misalignment Across Language Model Families
Jun 18, 2026Abdul Rafay SyedLLM Auditing Emergent Misalignment in Language Models
Department of Computer Science Universität des Saarlandes Saarbrücken, Germany
GEMS: Geometric Constraints Enable Multi-Semantic Superposition in LLMs
Jun 18, 2026Yu DengCausal Interventions in Language Models Representation Geometry in Language Models
Leverage Is Not Reach: A Control-Window Law for Single-Neuron Steering in Language Models
Jun 18, 2026Hongliang LiuLanguage Model Steering Causal Interventions in Language Models
Palo Alto Networks
Code-Switching Reveals Language Anchoring in Multilingual LLMs
Jun 18, 2026Jeonghyun Park, Seunghyun Yoon, Yonghyun Jun +1Multilingual Language Models Code-Switching
Chung-Ang University, Seoul, Korea · Adobe Research, USA
SAE Interventions are Unreliable: Post-Intervention Recovery of Suppressed Behavior
Jun 16, 2026Mingyue Cui, Linghui Shen, Xingyi YangSparse Autoencoders Mechanistic Interpretability
The Hong Kong Polytechnic University
A Closer Look at Failure Modes in Temporal Understanding of Large Audio-Language Models
Jun 16, 2026Apoorva Kulkarni, Kaousheik Jayakumar, Sreyan Ghosh +3Audio-Language Model Evaluation Audio-Language Models
University of Maryland, College Park, USA
Looking Is Not Picking: An Attention-Segment Account of Tool-Selection Failures in LLM Agents
Jun 15, 2026Shiyang ChenAI Agent Reliability Tool-Augmented Language Model Agents
Beijing Institute of Technology
Neuron Level Analysis of Large Language Model in Legal Domain Reasoning
Jun 14, 2026Eri Onami, Youmi Ma, Shuhei Kurita +1Legal NLP LLM Interpretability
Institute of Science Tokyo · NII · AIST
Vernier: Probing Representational Misalignment Behind Lexical Gaps in Causal Reasoning
Jun 14, 2026Zhenyu YuLLM Interpretability Causal Reasoning in Language Models
College of Computer Science and Artificial Intelligence Fudan University
Re-feeding Is Not Replaying: Measuring Replay Noise in Counterfactual Token-Credit Estimation
Jun 14, 2026Nils MattesonCounterfactual Evaluation Token-Level Credit Assignment
Northeastern University
Constitutional Value Potentials: reading and steering internal priority margins in language models
Jun 13, 2026Tong Che, Rui WuLanguage Model Safety Evaluation LLM Safety
Rutgers University
Beyond Importance: Interchange-Sobol Sensitivity Reveals Task-Specific Content Channels in Transformer Components
Jun 12, 2026Yifeng Guo, Jin-Hong Du, Xiang ChenMechanistic Interpretability Causal Interventions in Language Models
St. Jude Children’s Research Hospital Memphis, TN · The University of Hong Kong Hong Kong, CN
A Low-Rank Subspace Analysis of LLM Interventions
Jun 12, 2026Angira Sharma, Christian Schroeder de Witt, Philip Torr +2LLM Interpretability LLM Refusal Behavior
University of Oxford
TimeROME-DLM: Temporal Causal Tracing and Low-Rank Inference-Time Knowledge Editing for Masked Diffusion Language Models
Jun 11, 2026Zhengtao Yao, Liuyang Song, Hongbo Zhang +4Knowledge Editing Masked Diffusion Models
University of Southern California, USA · Peking University, China · Stevens Institute of Technology, USA +1
Can Editing 1 Neuron Fix Repetition Loops in LLMs?
Jun 9, 2026Aristotelis Lazaridis, Aman Sharma, Dylan Bates +3LLM Self-Correction Mechanistic Interpretability
Edgerunner AI
Steering Where to Listen: Instruction-Based Activation Steering Redirects Temporal Attention in Large Audio-Language Models
Jun 9, 2026Tsung-En Lin, Hung-Yi LeeLanguage Model Steering Audio Understanding
National Taiwan University, Taipei, Taiwan · NTU Artificial Intelligence Center of Research Excellence (NTU AI-CoRE)
From Observation to Intervention: A Causal Audit of Expert Importance in Mixture-of-Experts Models
Jun 9, 2026Leonard Engmann, Christian Medeiros Adriano, Holger GieseMixture-of-Experts Pruning LLM Auditing
Hasso Plattner Institute, University of Potsdam, Germany.
Causally Evaluating the Learnability of Formal Language Tasks
Jun 8, 2026Vésteinn Snæbjarnarson, Anej Svete, Josef Valvoda +3Language Modeling Causal Interventions in Language Models
1ETH Zürich · University of Copenhagen · *Now at Google
When Built-in Thinking Helps and Hurts: Constraint-Level Error Shifts in Instruction Following
Jun 8, 2026Sai Adith Senthil KumarLLM Evaluation Instruction Following
George Mason University
Emergent Misalignment Can Be Induced by Sycophancy and Reversed via Alignment Gating
Jun 8, 2026Sicheng Wang, Xiangyang Zhu, Han Wang +6LLM Sycophancy LLM Alignment
Shanghai AI Lab, Shanghai, China.
The Amplifying Mirror: Locating and Steering the Partisan Direction inside a Large Language Model
Jun 7, 2026Wendy K. TamLLM Interpretability Political Bias in Language Models
Professor and Stevenson Chair in the Departments of Political Science, Computer Science, Biomedical Informatics, and the Law School at Vanderbilt University
Pre-Intervention Prediction of Sparse Autoencoder Steering Side Effects
Jun 6, 2026Evan DuanLanguage Model Steering LLM Interpretability
Ablation-Reversible Heads Don't Transfer: A Stress Test for Mechanistic Role Claims in Transformers
Jun 6, 2026Philip QuirkeAttention Head Analysis Mechanistic Interpretability
Martian
Building Comparative Motivation Profiles with Instrumental Interventions
Jun 6, 2026David Vella Zarb, Rustem Turtayev, Taywon Min +2LLM Safety Evaluation Causal Interventions in Language Models
MATS · University of Cambridge · KAIST +1
When Behavioral Safety Evaluation Fails: A Representation-Level Perspective
Jun 6, 2026Enyi Jiang, Anders Gjølbye, Yibo Jacky Zhang +1Language Model Safety Evaluation LLM Auditing
1Stanford University · University of Illinois Urbana-Champaign · 3Technical University of Denmark
MechLens: Late Crystallization of Factual Knowledge Explains Intervention Effectiveness in Language Models
Jun 6, 2026Xueping GaoMemorization in Language Models LLM Interpretability
Alibaba Cloud
A Geometric Account of Activation Steering through Angle-Norm Decomposition
Jun 4, 2026Georgii Aparin, Tatiana GaintsevaLanguage Model Steering Language Modeling
Huawei Noah’s Ark Lab · Queen Mary University of London
The Piggyback Hypothesis of Generalization: Explaining and Mitigating Emergent Misalignment
Jun 4, 2026Jiachen Zhao, Zhengxuan Wu, Aryaman Arora +3LLM Alignment LLM Fine-Tuning
Northeastern University · University of California, Berkeley
The Self-Correction Illusion: LLMs Correct Others but Not Themselves
Jun 4, 2026Kuan-Yen Chen, Fang-Yi Su, Jung-Hsien ChiangLLM Self-Correction LLM Prompting
National Cheng Kung University

Jun 18, 2026, cs.LGJ/K move · Enter open · S save

Right Knowledge, Wrong Answer: Test-Time Steering for Temporal Fact Conflicts in Open-Weight Language Models

Elias Hossain, Sourav Saha, Umesh Chandra Biswas, Sanjeda Sara Jennifer

University of Central Florida · 2Mississippi State University

Knowledge Conflicts in Language Models Temporal Reasoning in Language Models Factual Knowledge in Language Models Causal Interventions in Language Models Activation Steering Temporal QA

Open paper AskPDF arXiv

Large language models can store both outdated facts and newer superseding facts in their parameters, but standard prompting may still elicit the outdated answer. We formalize this problem as Parametric Temporal Conflict (PTC) and introduce Temporal Attractor Steering (TAS), a three-stage test-time intervention that detects likely conflicts, identifies a conflict-critical layer, and steers hidden states toward newer-fact representations without retraining or external retrieval. We construct an 8,746-record verified benchmark across five Wikidata relations and evaluate four open-weight language models from three families: Qwen-2.5-1.5B/7B, Mistral-7B-v0.3, and Llama-3.1-8B. Single-layer activation patching achieves answer-flip rates of 0.72-0.85 across all models. End-to-end TAS resolves 29-57% of PTC cases while preserving 85-99% accuracy on non-conflict queries, outperforming a matched ITI baseline on three of four models. These results show that outdated parametric knowledge can be selectively overridden at inference time.

Momentum

Related topics

Latest papers 317