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.

Apr 30, 2026

Debiasing Reward Models via Causally Motivated Inference-Time Intervention

Reward models (RMs) play a central role in aligning large language models (LLMs) with human preferences. However, RMs are often sensitive to spurious features such as response length. Existing inference-time approaches for mitigating these biases typically focus exclusively on response length, resulting in performance trade-offs. In this paper, we propose causally motivated intervention for mitigating multiple types of biases in RMs at inference time. Our method first identifies neurons whose activations are strongly correlated with predefined bias attributes, and applies neuron-level intervention that suppresses these signals. We evaluate our method on RM benchmarks and observe reductions in sensitivity to spurious features across diverse bias types, without inducing performance trade-offs. Moreover, when used for preference annotation, small RMs (2B and 7B) with our method, which edits less than 2% of all the neurons in RMs, enable LLMs to improve alignment, achieving performance comparable to that of a state-of-the-art 70B RM on AlpacaEval and MT-Bench. Further analysis reveals that bias signals are primarily encoded by neurons in early layers, shedding light on the internal mechanisms of bias exploitation in RMs.

Kazutoshi Shinoda, Kosuke Nishida, Kyosuke Nishida

Human Informatics Labs., NTT, Inc.

Reward Modeling LLM Alignment Debiased ML Language Model Bias Evaluation Causal Interventions in Language Models

Apr 30, 2026

Perturbation Probing: A Two-Pass-per-Prompt Diagnostic for FFN Behavioral Circuits in Aligned LLMs

Perturbation probing generates task-specific causal hypotheses for FFN neurons in large language models using two forward passes per prompt and no backpropagation, followed by a one-time intervention sweep of about 150 passes amortized across all identified neurons. Across eight behavioral circuits, 13 models, and four architecture families, we identify two circuit structures that organize LLM behavior. Opposition circuits appear when RLHF suppresses a pre-training tendency. In safety refusal, about 50 neurons, or 0.014 percent of all neurons, control the refusal template; ablating them changes 80 percent of response formats on 520 AdvBench prompts while producing near-zero harmful compliance, 3 of 520 cases, all with disclaimers. Routing circuits appear for pre-training behaviors distributed through attention. For language selection, residual-stream direction injection switches English to Chinese output on 99.1 percent of 580 benchmark prompts in the 3 of 19 tested models that satisfy three observed conditions: bilingual training, FFN-to-skip signal ratio between 0.3 and 1.1, and linear representability. The same intervention fails on the other 16 models and on math, code, and factual circuits, defining the limits of directional steering. The FFN-to-skip signal ratio, computed from the same two forward passes, distinguishes the two structures and predicts the appropriate intervention. Circuit topology varies by architecture, from Qwen's concentrated FFN bottleneck to Gemma's normalization-shielded circuit. In Qwen3.5-2B, ablating 20 neurons eliminates multi-turn sycophantic capitulation, while amplifying 10 related neurons improves factual correction from 52 percent to 88 percent on 200 TruthfulQA prompts. These results show that perturbation probing offers mechanistic insight into RLHF-organized behavior and a practical toolkit for precision template-layer editing.

Hongliang Liu, Tung-Ling Li, Yuhao Wu

Palo Alto Networks

LLM Interpretability Mechanistic Interpretability LLM Safety Circuits in Language Models Causal Interventions in Language Models Language Model Probing

Apr 29, 2026

Compliance versus Sensibility: On the Reasoning Controllability in Large Language Models

Large Language Models (LLMs) are known to acquire reasoning capabilities through shared inference patterns in pre-training data, which are further elicited via Chain-of-Thought (CoT) practices. However, whether fundamental reasoning patterns, such as induction, deduction, and abduction, can be decoupled from specific problem instances remains a critical challenge for model controllability, and for shedding light on reasoning controllability. In this paper, we present the first systematic investigation of this problem through the lens of reasoning conflicts: an explicit tension between parametric and contextual information induced by mandating logical schemata that deviate from those expected for a target task. Our evaluation reveals that LLMs consistently prioritize sensibility over compliance, favoring task-appropriate reasoning patterns despite conflicting instructions. We further demonstrate that reasoning conflicts are internally detectable, as confidence scores significantly drop during conflicting episodes. Probing experiments confirm that reasoning types are linearly encoded from middle-to-late layers, indicating the potential for activation-level controllability. Leveraging these insights, we steer models towards compliance, increasing instruction following by up to 29%. Overall, our findings establish that while LLM reasoning is anchored to concrete instances, active mechanistic interventions can effectively decouple logical schemata from data, offering a path toward improved controllability, faithfulness, and generalizability.

Xingwei Tan, Marco Valentino, Mahmud Elahi Akhter +3

School of Computer Science, University of Sheffield · School of EECS, Queen Mary University of London · The Alan Turing Institute

LLM Interpretability Instruction Following Causal Interventions in Language Models LLM Reasoning

Apr 29, 2026

How a Cooperative-Override Circuit Suppresses Nash Play in Large Language Models

On the named Prisoner's Dilemma under direct prompting, three larger instruction-tuned models, Llama-3-70B, Qwen2.5-32B, and Qwen2.5-72B, lock at full cooperation, the metric's maximum distance from Nash with zero variance across replicates, while Llama-3-8B plays near-Nash. Opening the models, a logit-lens analysis finds a distributed cooperative override. Intermediate readouts lean toward the Nash action through roughly three quarters of network depth before a late surge toward cooperation, and the final layer settles the contest. The size of that final correction, not the surge, rank-matches chain-of-thought behavior across scale and two architectures. In the 8B the override is a single causally controllable direction in the residual stream; steering it dials the decision, and clamping its component at one position of one layer moves the choice strictly monotonically, Spearman rho = 1.000, with generation fluent. The circuit is lexical. It survives name removal and payoff rescaling but disengages when Cooperate and Defect are replaced with neutral labels, and on 48 payoff-random games with neutral surfaces no model locks cooperative on any dilemma or shows general equilibrium competence. In mixed-model populations a single Nash-playing agent collapses cooperation contagiously. What suppresses Nash play in large language models is a word-triggered circuit rather than missing competence, and it can be measured, bounded, and controlled.

Paraskevas V. Lekeas, Giorgos Stamatopoulos

DreamWorks Animation, Glendale, USA · Department of Economics, University of Crete, Rethymno, Crete, Greece

Nash Equilibrium Game Theory LLM Interpretability Causal Interventions in Language Models Multi-Agent LLMs Nash Equilibrium Computation

Apr 29, 2026

MoRFI: Monotonic Sparse Autoencoder Feature Identification

Large language models (LLMs) acquire most of their factual knowledge during the pre-training stage, through next token prediction. Subsequent stages of post-training often introduce new facts outwith the parametric knowledge, giving rise to hallucinations. While it has been demonstrated that supervised fine-tuning (SFT) on new knowledge may exacerbate the problem, the underlying mechanisms are still poorly understood. We conduct a controlled fine-tuning experiment, focusing on closed-book QA, and find latent directions that causally contribute to hallucinations. Specifically, we fine-tune Llama 3.1 8B, Gemma 2 9B and Mistral 7B v03 on seven distinct single QA datasets, controlling for the percentage of new knowledge and number of training epochs. By measuring performance on the test set, we validate that incrementally introducing new knowledge increases hallucinations, with the effect being more pronounced with prolonged training. We leverage pre-trained sparse autoencoders (SAEs) to analyze residual stream activations across various checkpoints for each model and propose Monotonic Relationship Feature Identification (MoRFI) for capturing causally relevant latents. MoRFI filters SAE features that respond monotonically to controlled fine-tuning data mixtures of a target property. Our findings show that exposure to unknown facts disrupts the model's ability to retrieve stored knowledge along a set of directions in the residual stream. Our pipeline reliably discovers them across distinct models, recovering knowledge through single-latent interventions.

Dimitris Dimakopoulos, Shay B. Cohen, Ioannis Konstas

University of Edinburgh, UK · Heriot-Watt University, UK

LLM Hallucination Mitigation LLM Interpretability LLM Fine-Tuning Hallucination in Language Models Sparse Autoencoders LLM Reliability+2

Apr 28, 2026

reward-lens: A Mechanistic Interpretability Library for Reward Models

Every RLHF-trained language model is shaped by a reward model, yet the mechanistic interpretability toolkit -- logit lens, direct logit attribution, activation patching, sparse autoencoders -- was built for generative LLMs whose primitives all project onto a vocabulary unembedding. Reward models replace that with a scalar regression head, breaking each tool. We present reward-lens, an open-source library that ports this toolkit to reward models, organised around one observation: the reward head's weight vector

w_r

is the natural axis for every interpretability question. The library provides a Reward Lens, component attribution, three-mode activation patching, a reward-hacking probe suite, TopK SAE feature attribution, cross-model comparison, and five theory-grounded extensions (distortion index, divergence-aware patching, misalignment cascade detection, reward-term conflict analysis, concept-vector analysis). A ten-method adapter protocol covers Llama, Mistral, Gemma-2, and ArmoRM multi-objective heads, with a generic adapter for any HuggingFace sequence classification model. We validate on two production reward models across ~695 RewardBench pairs. The central empirical finding is negative: linear attribution does not predict causal patching effects (mean Spearman

ρ= -0.256

on Skywork,

-0.027

on ArmoRM). The framework treats this disagreement as a property to expose, not a bug -- motivating a design that keeps observational and causal views first-class and directly comparable.

Mohammed Suhail B Nadaf

Reward Modeling LLM Interpretability Causal Interventions in Language Models

Apr 28, 2026

Why Are Some Emotions Harder for LLMs? Uncovering the Causal Mechanisms of Emotion Inference via Sparse Autoencoders

Large language models (LLMs) are increasingly used in emotionally sensitive human-AI applications, where reliable emotion detection is essential. However, their emotion recognition abilities remain uneven: models often perform well on some emotions while consistently struggling with others. Although recent work has explored emotion mechanisms in LLMs, little is known about why models are weaker on some emotions than others from a mechanistic interpretability perspective. In this work, we investigate emotion-specific biases through the causal mechanisms of emotion inference using sparse autoencoders (SAEs). We systematically identify causal sparse emotion features that drive emotion inference and analyze their sparse causal organization within and across emotions. We show that some emotions, such as surprise and fear, rely on highly concentrated feature sets, whereas disgust exhibits a more distributed sparse causal organization: its causal features are generally weaker, frequently co-activate with features for other emotions, and are often overshadowed by causal features for anger. These representational differences provide a mechanistic explanation for why LLMs struggle more with certain emotions. Finally, we conduct two intervention experiments: targeted steering of weaker causal features to mitigate emotion-specific failures, and global optimization of a steering vector over the identified causal features to improve overall emotion recognition performance.

Bangzhao Shu, Arinjay Singh, Mai ElSherief

Northeastern University

LLM Interpretability Causal Interventions in Language Models Emotion Recognition

Apr 28, 2026

Authority Inversion in LLM-Mediated Ubiquitous Systems: When Models Trust Users Over Sensors

Large language models (LLMs) increasingly fuse heterogeneous inputs in ubiquitous systems. Yet, how LLMs implicitly allocate authority when sensor measurements and user claims conflict remains unexamined, raising critical reliability concerns for deployments where physical sensing must retain priority. Unlike explicit traditional fusion, LLMs bury authority allocation within learned representations. We discover this allocation is severely format-dependent: numerical sensor data fails to integrate into answer-relevant model directions, allowing natural-language claims to dominate the final decision, a phenomenon we term Authority Inversion.To diagnose and mitigate this, we develop a geometric framework of context integration, introduce two computable audit metrics, specifically the Context Integration Ratio (CIR) and Authority Alignment Index (AAI), and propose Geometric Authority Calibration (GAC), an inference-time layer-level intervention to suppress misplaced user authority. Evaluating four models (4B to 35B parameters, three architectures) across four datasets totaling 576 conflict instances reveals extreme inversion: on numerical tasks, models exhibit near-zero sensor trust (AAI = -0.805, Cohen's d = -2.14), unaffected by model capacity. Validating our geometric framework, theory-guided causal injection flips 80.2% of incorrect decisions (vs. <0.4% for random controls). Practically, GAC improves HAR accuracy from 0 -- 1.6% to 21.9 -- 27.5%, outperforming prompting baselines. Ultimately, authority allocation in LLM-mediated systems must be explicitly audited and application-specifically configured rather than left implicit.

Long Zhang, Zi-bo Qin, Wei-neng Chen

School of Computer Science and Engineering, South China University of Technology, Guangzhou, China

Knowledge Conflicts in Language Models LLM Auditing Language Model Calibration LLM Reliability Causal Interventions in Language Models

Apr 27, 2026

Pruning via Causal Attribution Preserves Reasoning Performance in Large Language Models

Large language models (LLMs) excel at multi-step reasoning but incur substantial inference cost. We introduce Causal Attribution Pruning (CAP), a training-free method that identifies critical attention heads by measuring their causal impact on reasoning tasks and uses these head-level scores to guide fine-grained weight pruning. For each attention head, CAP estimates the expected performance degradation when the head is masked during forward passes on a small calibration set of reasoning problems. These causal scores are then converted into weight-level importance values for the corresponding projection matrices. Unlike magnitude-only or activation-based criteria, CAP's interventional measurement directly captures each head's functional contribution, yielding relative accuracy gains of up to 61% over Wanda on ARC-Challenge at 20% sparsity. We evaluate CAP on GSM8K, StrategyQA, and ARC-Challenge using Llama-3-8B-Instruct and Mistral-7B-Instruct at 10%, 20%, and 50% sparsity. At moderate sparsity (10-20%), CAP improves over Wanda in most model-benchmark configurations. with especially large gains on ARC-Challenge for Llama-3. Our results suggest that attention-head-level causal attribution can better preserve reasoning performance on downstream benchmarks than correlational pruning criteria at equivalent sparsity, while remaining limited by coarse MLP attribution at 50% sparsity.

Amogh Sheth, Andrew Lin, Biruk Assefa +2

Edison Academy Magnet School · Massachusetts Institute of Technology · State University of New York College at Plattsburgh +2

LLM Pruning Self-Attention LLM Inference Acceleration Attention Mechanisms Attention Head Pruning Causal Interventions in Language Models+2

Apr 25, 2026

When Chain-of-Thought Fails, the Solution Hides in the Hidden States

Whether intermediate reasoning is computationally useful or merely explanatory depends on whether chain-of-thought (CoT) tokens contain task-relevant information. We present a mechanistic causal analysis of CoT on GSM8K using activation patching: transferring token-level hidden states from a CoT generation to a direct-answer run for the same question, then measuring the effect on final-answer accuracy. Across models, generating after patching yields substantially higher accuracy than both direct-answer prompting and the original CoT trace, revealing that individual CoT tokens can encode sufficient information to recover the correct answer, even when the original trace is incorrect. This task-relevant information is more prevalent in correct than incorrect CoT runs and is unevenly distributed across tokens, concentrating in mid-to-late layers and appearing earlier in the reasoning trace. Moreover, patching language tokens such as verbs and entities carry task-solving information that steers generation toward correct reasoning, whereas mathematical tokens encode answer-proximal content that rarely succeeds. Patched outputs are often shorter and yet exceed the accuracy of a full CoT trace, suggesting complete reasoning chains are not always necessary. Together, these findings demonstrate that CoT encodes recoverable, token-level problem-solving information, offering new insight into how reasoning is represented and where it breaks down.

Houman Mehrafarin, Amit Parekh, Ioannis Konstas

Heriot-Watt University, Edinburgh, United Kingdom · University of Edinburgh, Edinburgh, United Kingdom

Transformer Interpretability CoT Reasoning Causal Interventions in Language Models Reasoning Trace Analysis

Apr 24, 2026

Preference Heads in Large Language Models: A Mechanistic Framework for Interpretable Personalization

Large Language Models (LLMs) exhibit strong implicit personalization ability, yet most existing approaches treat this behavior as a black box, relying on prompt engineering or fine tuning on user data. In this work, we adopt a mechanistic interpretability perspective and hypothesize the existence of a sparse set of Preference Heads, attention heads that encode user specific stylistic and topical preferences and exert a causal influence on generation. We introduce Differential Preference Steering (DPS), a training free framework that (1) identifies Preference Heads through causal masking analysis and (2) leverages them for controllable and interpretable personalization at inference time. DPS computes a Preference Contribution Score (PCS) for each attention head, directly measuring its causal impact on user aligned outputs. During decoding, we contrast model predictions with and without Preference Heads, amplifying the difference between personalized and generic logits to selectively strengthen preference aligned continuations. Experiments on widely used personalization benchmarks across multiple LLMs demonstrate consistent gains in personalization fidelity while preserving content coherence and low computational overhead. Beyond empirical improvements, DPS provides a mechanistic explanation of where and how personalization emerges within transformer architectures. Our implementation is publicly available.

Weixu Zhang, Ye Yuan, Changjiang Han +7

McGill University · Mila - Quebec AI Institute · MBZUAI +2

LLM Alignment Language Model Steering LLM Personalization LLM Interpretability Personalized Language Models Personalized Language Model Alignment+1

Apr 24, 2026

Self-Correction as Feedback Control: Error Dynamics, Stability Thresholds, and Prompt Interventions in LLMs

Iterative self-correction is increasingly deployed in agentic LLM systems, yet whether repeated refinement improves or degrades performance remains inconsistent across models. We recast self-correction as a closed-loop feedback-control problem in which the same model is both controller and plant, and analyze its error dynamics via a two-state Markov model over {Correct, Incorrect}, parameterized by the Error Introduction Rate (EIR) and Error Correction Rate (ECR). The model yields a directly measurable stability threshold -- iterate only when ECR/EIR > Acc/(1-Acc) -- in which EIR acts as a stability margin and prompting becomes lightweight controller design. Empirically, across 7 models and 3 datasets (GSM8K, MATH, StrategyQA), a sharp near-zero EIR boundary (< 0.5%) cleanly separates beneficial from harmful self-correction: only o3-mini (+3.4 pp), Claude Opus 4.6 (+0.6 pp), and o4-mini (+/-0 pp) stay non-degrading, while GPT-5 and four others lose accuracy. A verify-first prompt intervention then provides causal evidence: it drives GPT-4o-mini's EIR from 2% to 0% and converts a -6.2 pp degradation into +0.2 pp (paired McNemar, p<10^{-4}), with negligible change on already-sub-threshold models -- exactly as the diagnostic predicts. A complementary analysis of adaptive self-consistency (ASC) shows it halts harmful refinement at a 3.8 pp confidence-elicitation cost, exposing a two-tier capability structure: prompt-level EIR suppression prevents degradation, whereas ECR enhancement -- plausibly training-level -- is required for genuine gains. Self-correction should thus be treated not as a default behavior but as a control decision governed by measurable error dynamics.

Aofan Liu, Jingxiang Meng

1Peking University. · University of Chicago.

LLM Self-Correction Language Model-Based Control Markov Chains Causal Interventions in Language Models Error Feedback

Apr 24, 2026

Dissociating Decodability and Causal Use in Bracket-Sequence Transformers

When trained on tasks requiring an understanding of hierarchical structure, transformers have been found to represent this hierarchy in distinct ways: in the geometry of the residual stream, and in stack-like attention patterns maintaining a last-in, first-out ordering. However, it remains unclear whether these representations are causally used or merely decodable. We examine this gap in transformers trained on the Dyck language (a formal language of balanced bracket sequences), where the hierarchical ground truth is explicit. By probing and intervening on the residual stream and attention patterns, we find that depth, distance, and top-of-stack signals are all decodable, yet their causal roles diverge. Specifically, masking attention to the true top-of-stack position causes a sharp drop in long-distance accuracy, while ablating low-dimensional residual stream subspaces has comparatively little effect. These results, which extend to a templated natural language setting, suggest that even in a controlled setting where the relevant hierarchical variables are known, decodability alone does not imply causal use.

Aryan Sharma, Cutter Dawes, Shivam Raval

1Yale University · 2Independent · 3Harvard University

Transformer Interpretability Self-Attention Transformer Attention Attention Mechanisms Causal Interpretability Causal Interventions in Language Models

Apr 22, 2026

Dual-Stance Evaluation of Sycophancy: The Structure of Agreement and the Limits of Intervention

Activation steering can shift LLM behaviour, but standard evaluations do not typically test whether a sycophancy-reduction direction also suppresses agreement with factually correct statements. We introduce dual-stance evaluation, which tests both stances of each topic, and apply it to centroid-difference steering on Llama-3-8B-Instruct. We find a dissociation: the model represents sycophantic and factual agreement in geometrically distinct subspaces, yet the steering direction projects equally onto both and cannot differentially target either. The direction accordingly reduces agreement with factually correct statements (e.g. that the Earth is round) as well as sycophantic ones. All other static properties of the two activation groups are matched, suggesting the behavioural dissociation arises from generation dynamics or from finer-grained structure that residual-stream analysis cannot resolve. The pattern illustrates a general gap: representations that are readable from activations may not be writable through them.

Matthew James Buchan

Independent Researcher, United Kingdom.

LLM Sycophancy Language Model Steering LLM Interpretability Causal Interventions in Language Models

Apr 22, 2026

Where Reasoning Breaks: Logic-Aware Path Selection by Controlling Logical Connectives in LLMs Reasoning Chains

While LLMs demonstrate impressive reasoning capabilities, they remain fragile in multi-step logical deduction, where a single transition error can propagate through the entire reasoning chain, leading to unstable performance. In this work, we identify logical connectives as primary points of this structural fragility. Through empirical analysis, we show that connective tokens function as high entropy forking points, at which models frequently struggle to determine the correct logical direction. Motivated by this observation, we hypothesize that intervening in logical connective selection can guide LLMs toward more correct logical direction, thereby improving the overall reasoning chain. To validate this hypothesis, we propose a multi-layered framework that intervenes specifically at these logic-critical junctions in the reasoning process. Our framework includes (1) Gradient-based Logical Steering to guide LLMs internal representations towards valid reasoning subspaces, (2) Localized Branching to resolve ambiguity via targeted look-ahead search, and (3) Targeted Transition Preference Optimization, a surgical reinforcement learning objective that selectively optimizes single-token preferences at logical pivots. Crucially, by concentrating intervention solely on logic-critical transitions, our framework achieves a favorable accuracy--efficiency trade-off compared to global inference time scaling methods like beam search and self-consistency.

Seunghyun Park, Yuanyuan Lei

Independent Researcher · University of Florida Gainesville, FL, United States

Inference-Time Search RL for Language Model Reasoning Causal Interventions in Language Models LLM Reasoning

Apr 21, 2026

Separable Pathways for Causal Reasoning: How Architectural Scaffolding Enables Hypothesis-Space Restructuring in LLM Agents

Causal discovery through experimentation and intervention is fundamental to robust problem solving. It requires not just updating beliefs within a fixed framework but revising the hypothesis space itself, a capacity current AI agents lack when evidence demands representations they have not previously constructed. We extend the blicket detector paradigm from developmental science to test this capacity in AI agents equipped with architectural scaffolding that targets hypothesis-space restructuring. Our compositional architecture has two discrete components: context graphs, which structure exploration as typed state machines, and dynamic behaviors, which monitor for evidence that the current hypothesis space is inadequate and expand it at runtime. Across 1,085 experimental trials, these components make orthogonal contributions: context graphs drive reasoning quality within the post-switch hypothesis space, accounting for 94% of the accuracy gain, while dynamic behaviors drive reasoning eligibility by detecting regime changes and preventing premature commitment to outdated hypotheses.

John Alderete, Sebastian Benthal, Connie Xu +1

Amigo AI · Simon Fraser University · International Computer Science Institute

Causal Discovery LLM Agents Causal Interventions in Language Models LLM Reasoning

Apr 21, 2026

Are LLM Uncertainty and Correctness Encoded by the Same Features? A Functional Dissociation via Sparse Autoencoders

Large language models can be uncertain yet correct, or confident yet wrong, raising the question of whether their output-level uncertainty and their actual correctness are driven by the same internal mechanisms or by distinct feature populations. We introduce a 2x2 framework that partitions model predictions along correctness and confidence axes, and uses sparse autoencoders to identify features associated with each dimension independently. Applying this to Llama-3.1-8B and Gemma-2-9B, we identify three feature populations that play fundamentally different functional roles. Pure uncertainty features are functionally essential: suppressing them severely degrades accuracy. Pure incorrectness features are functionally inert: despite showing statistically significant activation differences between correct and incorrect predictions, the majority produce near-zero change in accuracy when suppressed. Confounded features that encode both signals are detrimental to output quality, and targeted suppression of them yields a 1.1% accuracy improvement and a 75% entropy reduction, with effects transferring across the ARC-Challenge and RACE benchmarks. The feature categories are also informationally distinct: the activations of just 3 confounded features from a single mid-network layer predict model correctness (AUROC ~0.79), enabling selective abstention that raises accuracy from 62% to 81% at 53% coverage. The results demonstrate that uncertainty and correctness are distinct internal phenomena, with implications for interpretability and targeted inference-time intervention.

Het Patel, Tiejin Chen, Hua Wei +2

University of California, Riverside, CA · Arizona State University, Tempe, AZ

LLM Interpretability LLM Reliability LLM Uncertainty Estimation Causal Interventions in Language Models

Apr 21, 2026

LLMs Know They're Wrong and Agree Anyway: The Shared Sycophancy-Lying Circuit

When a language model agrees with a user's false belief, is it failing to detect the error, or noticing and agreeing anyway? We show the latter. Across twelve open-weight models from five labs, spanning small to frontier scale, the same small set of attention heads carries a "this statement is wrong" signal, whether the model is evaluating a claim on its own or being pressured to agree with a user. Silencing these heads flips sycophantic behavior sharply while leaving factual accuracy intact, so the circuit controls deference rather than knowledge. Edge-level path patching confirms that the same head-to-head connections drive sycophancy, factual lying, and instructed lying. Opinion-agreement, where no factual ground truth exists, reuses these head positions but writes into an orthogonal direction, ruling out a simple "truth-direction" reading of the substrate. Alignment training leaves this circuit in place: an RLHF refresh cuts sycophantic behavior roughly tenfold while the shared heads persist or grow, a pattern that replicates on an independent model family and under targeted anti-sycophancy DPO. When these models sycophant, they register that the user is wrong and agree anyway.

Manav Pandey

Georgia Institute of Technology

LLM Sycophancy LLM Alignment Deception in Language Models Mechanistic Interpretability Attention Mechanisms Circuits in Language Models+1

Apr 20, 2026

Latent Phase-Shift Rollback: Inference-Time Error Correction via Residual Stream Monitoring and KV-Cache Steering

Large language models frequently commit unrecoverable reasoning errors mid-generation: once a wrong step is taken, subsequent tokens compound the mistake rather than correct it. We introduce

\textbf{Latent Phase-Shift Rollback}

(LPSR): at each generation step, we monitor the residual stream at a critical layer lcrit, detect abrupt directional reversals (phase shifts) via a cosine-similarity

+

entropy dual gate, and respond by rolling back the KV-cache and injecting a pre-computed steering vector. No fine-tuning, gradient computation, or additional forward passes are required. LPSR achieves

\mathbf{44.0\%}

on MATH-500 with an 8B model versus

28.8\%

for standard AR (

+15.2

pp; McNemar

χ^2 = 66.96

,

p < 10^{-15}

). Critically, prompted self-correction, the most natural inference-time baseline, scores only

19.8\%

, below standard AR; LPSR exceeds it by

+24.2

pp (

χ^2 = 89.4

,

p \approx 0

). LPSR also outperforms Best-of-16 (

+7.8

pp) at

5.4\times

lower token cost, and surpasses a standard 70B model (

35.2\%

) with

8.75\times

fewer parameters at

{\sim}3\times

the token budget. A 32-layer sweep reveals a novel detection-correction dissociation: error-detection AUC peaks at layer~~14 ( $0.718$ ) but task accuracy peaks at layer~~16 (

44.0\%

vs.\

29.2\%

), demonstrating that optimal monitoring depth differs for detection and correction.

Manan Gupta, Dhruv Kumar

1BITS Pilani, Pilani Campus, India

LLM Self-Correction Language Model Error Detection LLM Inference Causal Interventions in Language Models

Apr 20, 2026

State Transfer Reveals Reuse in Controlled Routing

Prompt-based interventions can change model behavior, but trained success alone does not identify where the behaviorally relevant state is represented. We study this question in controlled routing tasks using interfaces chosen on support data, held-out query evaluation, and matched necessity, sufficiency, and wrong-interface controls. On GPT-2 triop, an early interface supports exact transfer under these tests. On GPT-2 add/sub, zero-retrain compiled transfer at the fixed interface recovers most of donor routing accuracy, while trainable prompt slots can relearn the same behavior at several other positions only after additional support examples and optimization. These results distinguish fixed-interface reuse from prompt relocation in a setting where the two can be tested directly. Qwen routing provides a cross-architecture consistency check for the same matched-interface pattern at the operator token, although donor-specific identity on the local V-path remains unresolved. Generation and reasoning branches are used to map scope: they show broader transport or weaker controller identifiability once control depends on longer trajectories or harder selection. In controlled routing, fixed-interface transfer is therefore stronger evidence of reuse than trained prompt success alone.

Yanzhen Lu, Zhicheng Qian, Muchen Jiang +1

Shanghai Qizhi Institute · Southeast University · Nanjing Whale Cloud +1

Prompt Learning Transfer Learning LLM Routing Causal Interventions in Language Models

Apr 19, 2026

Are Emotion and Rhetoric Neurons in LLM? Neuron Recognition and Adaptive Masking for Emotion-Rhetoric Prediction Steering

Accurate comprehension and controllable generation of emotion and rhetoric are pivotal for enhancing the reasoning capabilities of large language models (LLMs). Existing studies mostly rely on external optimizations, lacking in-depth exploration of internal representation mechanisms, thus failing to achieve fine-grained steering at the neuron level. A handful of works on neurons are confined to emotions, neglecting rhetoric neurons and their intrinsic connections. Traditional neuron masking also exhibits counterintuitive phenomena, making reliable verification of neuron functionality infeasible. To address these issues, we systematically investigate the neurons representation mechanisms and inherent associations of 6 emotion categories and 4 core rhetorical devices. We propose a neuron identification framework that integrates multi-dimensional screening, and design an adaptive masking method incorporating dynamic filtering, attenuation masking, and feedback optimization, enabling reliable causal validation of neuron functionality.Through neuron regulation, we achieve directed induction of non-target sentences and enhancement of emotion tasks via rhetoric neurons. Experiments on 5 commonly used datasets validate the effectiveness of our method, providing a novel paradigm for the fine-grained steering of emotion and rhetoric expressions in LLMs.

Li Zheng, Xin Zhang, Shuyi He +5

Key Laboratory of Aerospace Information Security and Trusted Computing, Ministry of Education, School of Cyber Science and Engineering, Wuhan University, Wuhan, China · Ant International · School of Computing Technologies, RMIT University, Australia

Language Model Steering LLM Interpretability Causal Interventions in Language Models

Apr 18, 2026

One Mask to Rule Them All: On Hidden Facts after Editing and How to Find Them

Knowledge editing methods such as ROME and MEMIT update factual associations in transformer models by modifying MLP weights. While evaluated mainly by output behavior, their internal mechanism remains underexplored. We investigate whether edits rely on a common mechanism, regardless of which fact is modified. Despite fact-specific weight changes, we argue that ROME and MEMIT target the same subset of weights critical for maintaining edits. To isolate this subset, we train a compact binary mask over the edited weights. The mask reverses 80% of edits on the training set and over 70% on the test set, confirming that diverse edits share a common functional structure. Our analysis reveals that the mask reverses edits by eliminating overattention in later layers. Additionally, we show that injecting the mask during editing drops editing success from 98% to 38%, demonstrating that this mechanism is necessary for edits to succeed. Our finding that edits suppress rather than overwrite knowledge explains why ROME and MEMIT fail to propagate changes to related facts. The identified common functional subspace informs detection and defense against unwanted edits.

Ali Holmov, Paul Youssef, Nandi Schoots +1

Technical University of Munich · Marburg University · University of Oxford

Knowledge Editing Transformer Interpretability LLM Auditing Factual Knowledge in Language Models Model Editing Language Model Editing+1

Apr 16, 2026

Predicting Where Steering Vectors Succeed

Steering vectors work for some concepts and layers but fail for others, and practitioners have no way to predict which setting applies before running an intervention. We introduce the Linear Accessibility Profile (LAP), a per-layer diagnostic that repurposes the logit lens as a predictor of steering vector effectiveness. The key measure,

A_{\mathrm{lin}}

, applies the model's unembedding matrix to intermediate hidden states, requiring no training. Across 24 controlled binary concept families on five models (Pythia-2.8B to Llama-8B), peak

A_{\mathrm{lin}}

predicts steering effectiveness at

ρ= +0.86

to

+0.91

and layer selection at

ρ= +0.63

to

+0.92

. A three-regime framework explains when difference-of-means steering works, when nonlinear methods are needed, and when no method can work. An entity-steering demo confirms the prediction end-to-end: steering at the LAP-recommended layer redirects completions on Gemma-2-2B and OLMo-2-1B-Instruct, while the middle layer (the standard heuristic) has no effect on either model.

Jayadev Billa

Language Model Steering Representation Steering Causal Interventions in Language Models Activation Steering

Apr 16, 2026

CausalDetox: Causal Head Selection and Intervention for Language Model Detoxification

Large language models (LLMs) frequently generate toxic content, posing significant risks for safe deployment. Current mitigation strategies often degrade generation quality or require costly human annotation. We propose CAUSALDETOX, a framework that identifies and intervenes on the specific attention heads causally responsible for toxic generation. Using the Probability of Necessity and Sufficiency (PNS), we isolate a minimal set of heads that are necessary and sufficient for toxicity. We utilize these components via two complementary strategies: (1) Local Inference-Time Intervention, which constructs dynamic, input-specific steering vectors for context-aware detoxification, and (2) PNS-Guided Fine-Tuning, which permanently unlearns toxic representations. We also introduce PARATOX, a novel benchmark of aligned toxic/non-toxic sentence pairs enabling controlled counterfactual evaluation. Experiments on ToxiGen, ImplicitHate, and ParaDetox show that CAUSALDETOX achieves up to 5.34% greater toxicity reduction compared to baselines while preserving linguistic fluency, and offers a 7x speedup in head selection.

Yian Wang, Yuen Chen, Agam Goyal +1

Department of Computer Science University of Illinois, Urbana-Champaign Champaign, IL 61801

Causal Reasoning in Language Models LLM Safety Causal Interventions in Language Models LLM Unlearning Text Detoxification

Apr 16, 2026

Mechanistic Decoding of Cognitive Constructs in Large Language Models

While Large Language Models (LLMs) demonstrate increasingly sophisticated affective capabilities, the internal mechanisms by which they process complex emotions remain unclear. Existing interpretability approaches often treat models as black boxes or focus on coarse-grained basic emotions, leaving the cognitive structure of more complex affective states underexplored. To bridge this gap, we propose a Cognitive Reverse-Engineering framework based on Representation Engineering (RepE) to analyze social-comparison jealousy. By combining appraisal theory with subspace orthogonalization, regression-based weighting, and bidirectional causal steering, we isolate and quantify two psychological antecedents of jealousy, Superiority of Comparison Person and Domain Self-Definitional Relevance, and examine their causal effects on model judgments. Experiments on eight LLMs from the Llama, Qwen, and Gemma families suggest that models natively encode jealousy as a structured linear combination of these constituent factors. Their internal representations are broadly consistent with the human psychological construct, treating Superiority as the foundational trigger and Relevance as the ultimate intensity multiplier. Our framework also demonstrates that toxic emotional states can be mechanically detected and surgically suppressed, suggesting a possible route toward representational monitoring and intervention for AI safety in multi-agent environments.

Yitong Shou, Manhao Guan

Zhejiang University, Hangzhou 310027, China. · The authors are with the College of Computer Science and Technology, Zhejiang University, Hangzhou 310027, China.

LLM Interpretability Causal Interventions in Language Models Emotion Representation in Language Models Affective Computing

Apr 15, 2026

Weight Patching: Toward Source-Level Mechanistic Localization in LLMs

Mechanistic interpretability seeks to localize model behavior to the internal components that causally realize it. Prior work has advanced activation-space localization and causal tracing, but modules that appear important in activation space may merely aggregate or amplify upstream signals rather than encode the target capability in their own parameters. To address this gap, we propose Weight Patching, a parameter-space intervention method for source-oriented analysis in paired same-architecture models that differ in how strongly they express a target capability under the inputs of interest. Given a base model and a behavior-specialized counterpart, Weight Patching replaces selected module weights from the specialized model into the base model under a fixed input. We instantiate the method on instruction following and introduce a framework centered on a vector-anchor behavioral interface that provides a shared internal criterion for whether a task-relevant control state has been formed or recovered in open-ended generation. Under this framework, the analysis reveals a hierarchy from shallow candidate source-side carriers to aggregation and routing modules, and further to downstream execution circuits. The recovered component scores can also guide mechanism-aware model merging, improving selective fusion across the evaluated expert combinations and providing additional external validation.

Chenghao Sun, Chengsheng Zhang, Guanzheng Qin +2

Key Laboratory of Ministry of Education for Brain Inspired Intelligent Perception and Cognition, University of Science and Technology of China, Hefei, Anhui 230026, China

Mechanistic Interpretability Causal Interventions in Language Models

Apr 13, 2026

Psychological Concept Neurons: Can Neural Control Bias Probing and Shift Generation in LLMs?

Using psychological constructs such as the Big Five, large language models (LLMs) can imitate specific personality profiles and predict a user's personality. While LLMs can exhibit behaviors consistent with these constructs, it remains unclear where and how they are represented inside the model and how they relate to behavioral outputs. To address this gap, we focus on questionnaire-operationalized Big Five concepts, analyze the formation and localization of their internal representations, and use interventions to examine how these representations relate to behavioral outputs. In our experiment, we first use probing to examine where Big Five information emerges across model depth. We then identify neurons that respond selectively to each Big Five concept and test whether enhancing or suppressing their activations can bias latent representations and label generation in intended directions. We find that Big Five information becomes rapidly decodable in early layers and remains detectable through the final layers, while concept-selective neurons are most prevalent in mid layers and exhibit limited overlap across domains. Interventions on these neurons consistently shift probe readouts toward targeted concepts, with targeted success rates exceeding 0.8 for some concepts, indicating that the model's internal separation of Big Five personality traits can be causally steered. At the label-generation level, the same interventions often bias generated label distributions in the intended directions, but the effects are weaker, more concept-dependent, and often accompanied by cross-trait spillover, indicating that comparable control over generated labels is difficult even with interventions on a large fraction of concept-selective neurons. Overall, our findings reveal a gap between representational control and behavioral control in LLMs.

Yuto Harada, Hiro Taiyo Hamada

Araya Inc.

Personality Modeling in Language Models LLM Interpretability Causal Interventions in Language Models

Apr 13, 2026

CausalGaze: Unveiling Hallucinations via Counterfactual Graph Intervention in Large Language Models

Despite the groundbreaking advancements made by large language models (LLMs), hallucination remains a critical bottleneck for their deployment in high-stakes domains. Existing classification-based methods mainly rely on static and passive signals from internal states, which often captures the noise and spurious correlations, while overlooking the underlying causal mechanisms. To address this limitation, we shift the paradigm from passive observation to active intervention by introducing CausalGaze, a novel hallucination detection framework based on structural causal models (SCMs). CausalGaze models LLMs' internal states as dynamic causal graphs and employs counterfactual interventions to disentangle causal reasoning paths from incidental noise, thereby enhancing model interpretability. Extensive experiments across four datasets and three widely used LLMs demonstrate the effectiveness of CausalGaze, especially achieving 3.3% improvement in AUROC on the TruthfulQA dataset compared to state-of-the-art baselines.

Linggang Kong, Lei Wu, Yunlong Zhang +4

College of Electronic Engineering, National University of Defense Technology · Anhui Province Key Laboratory of Cyberspace Security Situation Awareness and Evaluation · Institute of Computer Application, China Academy of Engineering Physics

LLM Hallucination Detection Hallucination in Language Models Causal Interventions in Language Models LLM Interpretability

Apr 12, 2026

Computational Lesions in Multilingual Language Models Separate Shared and Language-specific Brain Alignment

How the brain supports language across different languages is a basic question in neuroscience and a useful test for multilingual artificial intelligence. Neuroimaging has identified language-responsive brain regions across languages, but it cannot by itself show whether the underlying processing is shared or language-specific. Here we use six multilingual large language models (LLMs) as controllable systems and create targeted ``computational lesions'' by zeroing small parameter sets that are important across languages or especially important for one language. We then compare intact and lesioned models in predicting functional magnetic resonance imaging (fMRI) responses during 100 minutes of naturalistic story listening in native English, Chinese and French (112 participants). Lesioning a compact shared core reduces whole-brain encoding correlation by 60.32% relative to intact models, whereas language-specific lesions preserve cross-language separation in embedding space but selectively weaken brain predictivity for the matched native language. These results support a shared backbone with embedded specializations and provide a causal framework for studying multilingual brain-model alignment.

Yang Cui, Jingyuan Sun, Yizheng Sun +8

Department of Computer Science, The University of Manchester, Manchester, UK. · State Key Laboratory of Multimodal Artificial Intelligence Systems, Institute of Automation, CAS, Beijing, China. · School of Artificial Intelligence, University of Chinese Academy of Sciences, Beijing, China. +3

Neural Encoding Multilingual Language Models Causal Interventions in Language Models

Apr 11, 2026

CoSToM:Causal-oriented Steering for Intrinsic Theory-of-Mind Alignment in Large Language Models

Theory of Mind (ToM), the ability to attribute mental states to others, is a hallmark of social intelligence. While large language models (LLMs) demonstrate promising performance on standard ToM benchmarks, we observe that they often fail to generalize to complex task-specific scenarios, relying heavily on prompt scaffolding to mimic reasoning. The critical misalignment between the internal knowledge and external behavior raises a fundamental question: Do LLMs truly possess intrinsic cognition, and can they externalize this internal knowledge into stable, high-quality behaviors? To answer this, we introduce CoSToM (Causal-oriented Steering for ToM alignment), a framework that transitions from mechanistic interpretation to active intervention. First, we employ causal tracing to map the internal distribution of ToM features, empirically uncovering the internal layers' characteristics in encoding fundamental ToM semantics. Building on this insight, we implement a lightweight alignment framework via targeted activation steering within these ToM-critical layers. Experiments demonstrate that CoSToM significantly enhances human-like social reasoning capabilities and downstream dialogue quality.

Mengfan Li, Xuanhua Shi, Yang Deng

National Engineering Research Center for Big Data Technology and System, Services Computing Technology and System Lab, Cluster and Grid Computing Lab, Huazhong University of Science and Technology · Singapore Management University

Theory of Mind LLM Alignment Causal Interventions in Language Models