Causal Interventions in Language Models

Follow this topic Ask about this topic

This field asks which internal states, features, heads, neurons, or context elements actually influence a language model’s decisions, rather than merely correlate with them. Researchers use counterfactual prompt and evidence edits, activation patching and ablation, causal tracing, sparse-autoencoder features, and inference-time steering or weight edits to test mechanisms and control behavior. Applications include checking whether chain-of-thought and retrieved evidence drive answers, and modifying refusal, hallucination, bias, tool use, or reasoning while preserving unrelated capabilities. Results show that readable representations are often not causally used or easily steered; effects vary by layer, task, model, and context, while redundancy and collateral effects make reliable, selective interventions a continuing challenge.

Latest papers

489 papers83 papers in the last four weeks, up 22% on the four weeks before. 0.5% of all new papers.

Function-Vector Heads Are Two Populations: Writers and Cancellers in In-Context Learning
May 25, 2026Han-yu WangTransformer Interpretability Task Vectors
The University of Hong Kong
Priors Persist Through Suppression: A Stroop Paradigm for Lexical Override
May 25, 2026Han-yu WangKnowledge Conflicts in Language Models LLM Interpretability
The University of Hong Kong
CausalFlow: Causal Attribution and Counterfactual Repair for LLM Agent Failures
May 25, 2026Akash Bonagiri, Devang Borkar, Gerard Janno Anderias +2Agent Failure Analysis Causal Interventions in Language Models
Department of Computer Science University of California, Davis Davis, CA 95616
Transformer Field Theory: A Response-Theoretic Approach to Mechanistic Interpretability
May 24, 2026David N. Olivieri, Antonio F. Pérez RodríguezTransformer Interpretability Mechanistic Interpretability
Departamento de Informática, Universidade de Vigo, Spain · Independent Researcher, Spain
Representation Without Control: Testing the Realization Effect in Language Models
May 24, 2026Ciarán Walsh, Emilio BarkettLLM Interpretability Causal Interventions in Language Models
Columbia University
Faithfulness as Information Flow: Evaluating and Training Faithful Chain-of-Thought Reasoning
May 22, 2026Jinghan Jia, Joe Benton, Eric EasleyCoT Faithfulness LLM Interpretability
†Dept. CSE, Michigan State University · ⋆Anthropic Fellows Program · ‡Anthropic
As X, Do Y: How Persona and Task Combine in Instruction-Tuned LLMs
May 22, 2026Eric XuLLM Prompting LLM Interpretability
Independent Researcher
Represented Is Not Computed: A Causal Test of Candidate Algorithmic Intermediates in a Transformer
May 21, 2026Ishita Darade, Sushrut ThoratNumerical Reasoning in Language Models Transformer Interpretability
MKSSS’s Cummins College of Engineering for Women, Pune 411052, India · Institute of Cognitive Science, Osnabrück University, Osnabrück 49069, Germany
From Correlation to Cause: A Five-Stage Methodology for Feature Analysis in Transformer Language Models
May 21, 2026Caleb MunigetyTransformer Interpretability Causal Reasoning in Language Models
Independent Researcher
Diagnosis Is Not Prescription: Linguistic Co-Adaptation Explains Patching Hazards in LLM Pipelines
May 21, 2026Yoon Jeonghun, Kim DongchanLLM Agents Causal Interventions in Language Models
KAIST (Korea Advanced Institute of Science and Technology), Republic of Korea · NAVER Corp., Seongnam, Republic of Korea · NAVER Corp., Bellevue, WA, USA
Causal Evidence for Attention Head Imbalance in Modality Conflict Hallucination
May 19, 2026Jinrui Jiang, Zhangtai Wu, Zhen Wu +1Hallucination in Language Models Multimodal Large Language Models
National Key Laboratory for Novel Software Technology, Nanjing University, China · School of Artificial Intelligence, Nanjing University, China
Counterfactual Likelihood Tests for Indirect Influence in Private Reasoning Channels
May 18, 2026Alexander Boesgaard LorupCounterfactual Evaluation LLM Interpretability
Openhagen
TRACE: Trajectory Correction from Cross-layer Evidence for Hallucination Reduction
May 18, 2026Tej Sanibh RanadeLLM Hallucination Mitigation Hallucination in Language Models
Independent Researcher
Causal Intervention-Based Memory Selection for Long-Horizon LLM Agents
May 17, 2026Saksham Sahai SrivastavaConversational Memory Long-Term Conversational Memory
School of Computing, University of Georgia, Athens, Georgia, USA.
The Point of No Return: Counterfactual Localization of Deceptive Commitment in Language-Model Reasoning
May 16, 2026Scott Merrill, Shashank SrivastavaLanguage Model Safety Evaluation LLM Interpretability
University of North Carolina at Chapel Hill
To Call or Not to Call: Diagnosing Intrinsic Over-Calling Bias in LLM Agents
May 16, 2026Wei Shi, Ziheng Peng, Sihang Li +4Tool-Augmented Language Model Agents Causal Interventions in Language Models
Shanghai Jiao Tong University · Shanghai Artificial Intelligence Laboratory · Renmin University of China +2
Artificial Aphasias in Lesioned Language Models
May 15, 2026Nathan Roll, Jill Kries, Laura Gwilliams +1LLM Interpretability Causal Interventions in Language Models
Department of {1Linguistics, 2Psychology}, Stanford University · 3Wu Tsai Neurosciences Institute, Stanford University
Non-linear Interventions on Large Language Models
May 14, 2026Sangwoo KimLLM Interpretability Causal Interventions in Language Models
Department of Linguistics, Seoul National University, Republic of Korea.
Does RAG Know When Retrieval Is Wrong? Diagnosing Context Compliance under Knowledge Conflict
May 14, 2026Yihang Chen, Pin Qian, Su Wang +4Retrieval-Augmented Generation Knowledge Conflicts in Language Models
1Georgia Institute of Technology · 2Carnegie Mellon University · University of California San Diego
Faithful or Fabricated? A Causal Framework for Rationalization Bias in LLM Judges
May 13, 2026Riya Tapwal, Abhishek Kumar, Carsten MapleLLM-as-a-Judge Faithfulness of Language Model Explanations
Indian Institute of Technology (IIT) Mandi · The Alan Turing Institute, London, U.K. · Warwick Manufacturing Group University of Warwick, U.K.
Not Just RLHF: Why Alignment Alone Won't Fix Multi-Agent Sycophancy
May 13, 2026Adarsh Kumarappan, Ananya MujooAI Agent Reliability Multi-Agent LLM Systems
1California Institute of Technology · 2Evergreen Valley College
When Attention Closes: How LLMs Lose the Thread in Multi-Turn Interaction
May 13, 2026Vardhan Dongre, Joseph Hsieh, Viet Dac Lai +3Self-Attention LLM Interpretability
University of Illinois Urbana-Champaign · Adobe Research
WriteSAE: Sparse Autoencoders for Recurrent State
May 12, 2026Jack YoungRecurrent Language Models Language Model Steering
Indiana University
Fair outputs, Biased Internals: Causal Potency and Asymmetry of Latent Bias in LLMs for High-Stakes Decisions
May 12, 2026Jagdish Tripathy, Marcus BuckmannLLM Auditing Implicit Bias
Bank of England
Temporal Preference Concepts and their Functions in a Large Language Model
May 11, 2026Ian Rios-Sialer, Shantanu Darveshi, Shuai Jiang +4LLM Interpretability Temporal Reasoning in Language Models
AISC San Francisco, USA · AISC Munich, Germany · SPAR, AISC Berlin, Germany
The Last Word Often Wins: A Format Confound in Chain-of-Thought Corruption Studies
May 11, 2026Gabriel GarciaCoT Faithfulness CoT Evaluation
Independent Researcher
Prompt-Activation Duality: Improving Activation Steering via Attention-Level Interventions
May 11, 2026Diancheng Kang, Zheyuan Liu, Ningshan Ma +3Language Model Steering Self-Attention
1Southern University of Science and Technology · University of Notre Dame · 3Massachusetts Institute of Technology
Vocabulary Hijacking in LVLMs: Unveiling Critical Attention Heads by Excluding Inert Tokens to Mitigate Hallucination
May 11, 2026Yangneng Chen, Junlin Li, Weijun Yao +4Hallucination in Language Models VLM Hallucination
1Harbin Institute of Technology (Shenzhen), China · 2Huawei Technologies Co., Ltd. · 3The Hong Kong Polytechnic University +1
Exploitation Without Deception: Dark Triad Feature Steering Reveals Separable Antisocial Circuits in Language Models
May 10, 2026Cameron Berg, Roshni LullaLLM Interpretability Personality Modeling in Language Models
1Reciprocal Research · 2Brain & Creativity Institute, University of Southern California
Causal state binding predicts action control in language agents
May 10, 2026Xiao JiaLanguage Model-Based Control LLM Agent Evaluation
School of Artificial Intelligence, The Chinese University of Hong Kong, Shenzhen

May 25, 2026J/K move · Enter open · S save

Function-Vector Heads Are Two Populations: Writers and Cancellers in In-Context Learning

Han-yu Wang

The University of Hong Kong

Transformer Interpretability Task Vectors In-Context Learning Causal Interventions in Language Models Function Vectors Attention Head Specialization

Open paper AskPDF arXiv

In-context learning lets a language model perform a task specified by examples in its prompt. Function vectors capture task information in a compact activation assembled from attention-head outputs. Across two rule families and three Pythia models, we find two opposed functional populations among candidate function-vector heads. Writers support the rule-correct answer, while cancellers systematically suppress it. These roles predict opposing group effects on held-out prompts, where removing cancellers improves accuracy by 2.4 to 6.8 percentage points in all six settings. The dominant suppressive contributions depend on task content, and the same components can support correct predictions in another task. At the population level, substantial support and suppression can nearly cancel. Separating these roles explains how task-related computation can work against correct predictions and how a small aggregate effect can conceal strong opposing contributions.