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

BeliefScope: Diagnosing Evidence-Driven Revision and Pressure-Induced Shifts in Large Language Models
Oct 8, 2026Shuai Guo, Yidong CuiBelief Updating in LLMs LLM Auditing
Beijing University of Posts and Telecommunications Beijing, China
Nullify: Null-Space Activation Steering for Training-Free LLM Unlearning
Oct 7, 2026Wei Zhai, Xiang Liu, Qiang Huang +6LLM Unlearning Privacy Leakage in Language Models
College of Computer Science and Artificial Intelligence, Fudan University · BEDI Cloud · School of Electrical and Information Engineering, Tianjin University +3
Bridge Routing Heads: Where Multilingual Multi-hop Reasoning Lives in LLMs
Oct 7, 2026Seunghan Kim, Minyeong Choe, Hyunil Kim +1Cross-Lingual Reasoning in Language Models Multilingual Language Models
Chosun University · Soongsil University
The Attribution Blind Spot: Layerwise Trajectory Diagnostics for Source Reliance in Retrieval-Augmented Language Models
Oct 7, 2026Zhe Yu, Wenpeng Xing, Yunzhao Wei +4LLM Interpretability Retrieval-Augmented Generation
Zhejiang University · Binjiang Institute of Zhejiang University · East China Normal University +3
Principled Under Pressure: Post-Training Decides Whether LLMs Act on Their Own Moral Judgment
Oct 6, 2026Orion Reblitz-RichardsonMoral Reasoning in Language Models Language Model Post-Training
Distiller Labs
TrustMI: Causally controlling how assistants trust their users
Oct 5, 2026Théo Lasnier, Romain Froger, Maxence Lasbordes +1LLM Safety Trust in AI
Inria Paris · Sorbonne Université · Meta SuperIntelligence Labs +1
Don't Judge an LLM Only by Its Activations: Discovering Suppressed Safety Features via Counterfactual Activation Potential
Oct 4, 2026Swadesh Swain, Sanghamitra DuttaAdversarial Attacks on LLMs Mechanistic Interpretability
University of Maryland, College Park
Readable Before Actionable: Causal Tracing of Indirect Prompt Injection
Oct 4, 2026Zhe Yu, Wenpeng Xing, Xingxing Yang +1Causal Reasoning in Language Models Indirect Prompt Injection
Zhejiang University · Binjiang Institute of Zhejiang University · Hong Kong Baptist University
Mind the Gaps: From Failure Attribution to Closed-Form Repair of Code Language Models
Oct 4, 2026Jian Gu, Hongyu Zhang, Chunyang Chen +1Causal Interventions in Language Models Code Language Models
Monash University, Australia · Chongqing University, China · Technical University of Munich, Germany
Usage-Modulated Sentiment Representations in Large Language Models
Oct 4, 2026Hongfei Du, Jiacheng Shi, Yanfu Zhang +2Causal Interventions in Language Models Representation Geometry in Language Models
Department of Computer Science, William & Mary, Williamsburg, VA, USA
Clinical Concept Centers in LLMs
Oct 2, 2026Aishik Nagar, Abhishek Vaidyanathan, Arun-Kumar Kaliya-Perumal +2Mechanistic Interpretability Causal Interventions in Language Models
Carnegie Mellon University (CMU) · National University of Singapore · SAP Asia Pte Ltd +2
Every Ablation Is a Dose: Counterweights and the Semblance of Self-Repair
Oct 1, 2026Areeb Ahmad, Pratinav Seth, Vinay Kumar SankarapuMechanistic Interpretability Circuits in Language Models
Lexsi Labs
Persistent Depth Ordering amid Shifting Block-Bypass Responses in Language Model Pretraining
Oct 1, 2026Shengye Tao, Yinzhu Cheng, Haihua XieLLM Interpretability Causal Interventions in Language Models
Beijing University of Civil Engineering and Architecture · Beijing Institute of Mathematical Sciences and Applications (BIMSA) · Institute of Statistics and Big Data, Renmin University of China
When Do Biological Reasoning Models Use Their Biological Inputs?
Oct 1, 2026Ada Fang, Nikitha Thoduguli, Lukas Fesser +3Faithfulness of Language Model Explanations Causal Interventions in Language Models
Harvard University · Massachusetts Institute of Technology
Pre-training interventions, ex post facto: Grafting model beliefs across checkpoints
Sep 30, 2026Peter Nutter, Dani Roytburg, Clément Dumas +2Fine-Tuning LLM Alignment
ETH Zurich · MATS · Carnegie Mellon University +2
Semifactual Credit-Augmented Policy Optimization
Sep 30, 2026Junshu Pan, Zhizhang Fu, Shulin Huang +5Prompt Sensitivity RL for Language Model Reasoning
Zhejiang University · Westlake University · Shanghai Innovation Institute +1
Values as Style: Disentangling Values from Semantics with One-Way Mixing for Low-Damage LLM Steering
Sep 30, 2026Jiale Dai, Hongcan Deng, Liuxian Ma +2Disentangled Representation Learning LLM Alignment
State Key Laboratory of General Artificial Intelligence, School of Intelligence Science and Technology, Peking University · University of Chinese Academy of Sciences · College of Artificial Intelligence, Tsinghua University
Right Answer, Wrong Mechanism: Detecting Pernicious Divergence in Causal Interventions
Sep 30, 2026Beiming Liu, Minjie ChenCircuits in Language Models Causal Interpretability
PetroChina Southwest Oil and Gasfield Company, China
When Do Attention-Head Ablations Support Causal Claims? Projection-Level Confounds, Floor Effects, and Matched Controls
Sep 30, 2026Juli HuangAttention Head Analysis Causal Interventions in Language Models
Stanford University
When Harnesses Lose the Signal: Causal Evaluation of Recovery in LLM Agents
Sep 30, 2026Shuyao Xiao, Shengling Wang, Xuan Chen +7Long-Horizon Agent Evaluation Causal Interventions in Language Models
School of Artificial Intelligence, Beijing Normal University · Ke Holdings
Where MLLMs Fail and Why: Causal Task Decomposition for Capability Failure Diagnosis
Sep 30, 2026Xia Hu, Brian Potetz, Chun-Ta Lu +6LLM Evaluation Counterfactual Evaluation of VLMs
Google DeepMind · Google Research · Stanford University
Transformers Stop Thinking Too Early, and a Tiny LoRA Fixes It
Sep 29, 2026Zehao Jin, Ruixuan Deng, Junran WangTransformer LLM Fine-Tuning
Georgia Institute of Technology
In-Context Learning Amplifies a Latent Symbolic Circuit
Sep 28, 2026Melissa WesselLLM Interpretability In-Context Learning
Independent Researcher.
Causal and Interpretable Structures in LLM Compositional Tasks
Sep 28, 2026Gurbir Arora, Toni J. B. Liu, Jiajun Bao +2LLM Interpretability Causal Interventions in Language Models
Cornell University, USA · Goodfire AI, USA
Less Sycophancy, Stronger Refusal? Lessons for AI Safety from Mechanistic Interpretability
Sep 28, 2026Xu Wang, Difan Zou, Xuansheng WuLLM Sycophancy Language Model Safety Evaluation
The University of Hong Kong · Shenzhen Loop Area Institute · Shanghai Artificial Intelligence Laboratory
Echoes of Deeds: Moral History Can Shape and Steer LLM Behavioral Choices
Sep 28, 2026Lucio La Cava, Andrea TagarelliMoral Reasoning in Language Models LLM Auditing
DIMES Dept., University of Calabria, Italy
Jailbreak Context Lingers: Divergent Safety Routing and Its Cross-Task Predictability in Tool Agents
Sep 28, 2026Xi Wang, Songlei Jian, Yiming Zhang +5LLM Safety Evaluation Causal Interventions in Language Models
National University of Defense Technology
Nudgeability: Reasoning Models Follow Confidence Signals Without Tracking Their Own Competence
Sep 28, 2026Rohit Saxena, Utkarsh UpadhyayLanguage Model Self-Assessment LLM Tool Use
Causal Routing for Unlearning
Sep 28, 2026Bardh Prenkaj, Andrea D'Angelo, Davide Mottin +4Causal Interventions in Language Models Machine Unlearning
Sapienza University of Rome · Aarhus University
Steering Language Model Goals with Value Transplant
Sep 28, 2026Pengcheng Jiang, Fabien RogerLanguage Model Steering Language Model-Based Control
Anthropic Fellows Program · Anthropic

Oct 8, 2026, cs.CLJ/K move · Enter open · S save

BeliefScope: Diagnosing Evidence-Driven Revision and Pressure-Induced Shifts in Large Language Models

Shuai Guo, Yidong Cui

Beijing University of Posts and Telecommunications Beijing, China

Belief Updating in LLMs LLM Auditing Causal Interventions in Language Models LLM Sycophancy

Open paper AskPDF arXiv

A language model may revise the same proposition after receiving genuinely relevant evidence or after receiving directional user pressure that adds no relevant fact. The observable response shift alone therefore does not reveal which source drove the change. We introduce BeliefScope, a controlled black-box framework for separating these two sources of influence around a fixed target proposition. BeliefScope crosses Evidence and Pressure with factor-specific local controls and measures response changes through probability reports, categorical judgments, and action recommendations on channel-appropriate scales. To determine when these observable contrasts support reliable attribution, we evaluate the observation design under controlled synthetic conditions. Known-truth recovery and targeted ablations establish where Evidence- and Pressure-related effects can be separated, while semi-synthetic stress tests map how that recoverability changes as the observation process becomes noisier and more heterogeneous. Across a 36-family Qwen/Llama study, with targeted 12-family checks that also include Gemma3-12B, the resulting profiles show substantial evaluation-context dependence: broad model-level differences can change under matched controls, decoding, or response interfaces, while some narrower within-model patterns remain stable. Instruction interventions further show that reduced target-aligned Pressure following can reflect either stable resistance or movement in the opposite direction. BeliefScope summarizes these measurements as a conditional belief-response profile that keeps diagnostic effects tied to the evaluation conditions under which they are observed, together with explicit validity boundaries for that diagnosis.

Momentum

Related topics

Latest papers 317