MechInterp @ NeurIPS

136 papers82 directions

Papers by Year

2025 136

Top Research Directions

Research directions most represented by papers from this venue.

Resolving Superposition with Sparse Autoencoders 8
Probing the Internal Representations of Large Language Models 7
Understanding Feature Superposition in Deep Neural Networks 5
Activation Steering for Inference Time Model Control 5
Projecting Intermediate Transformer States to the Vocabulary Space 3
Activation Patching for Mechanistic Interpretability 3
Steering Models with Contrastive Activation Vectors 2
Scaling Mechanistic Interpretability to Large Language Models 2
Localized Parameter Editing for Large Language Models 2
Linear Probing of Internal Network Representations 2
The Cross Architecture Validation Gap in Foundation Models 2
Generalizing Machine Learning Frameworks Across Network Architectures 2
Evaluating the Faithfulness of Machine Learning Explanations 2
The Causal Disconnect in Language Model Reasoning 2
Scaling Specialized Language Model Interventions to Frontier Models 2

Top Authors

AuthorPapers
Neel Nanda8
Kevin Zhu6
Constantin Venhoff5
Samuel Marks5
Arthur Conmy4
Ashwinee Panda4
Ge Yan4
Sean O'Brien4
Senthooran Rajamanoharan4
Tsui-Wei Weng4
Adam Karvonen3
Adam Stein3
Christian Schroeder de Witt3
David Bau3
Edward Stevinson3
Eric Wong3
Gonçalo Paulo3
Nora Belrose3
Philip Torr3
Tamar Rott Shaham3

Impactful Papers

Papers from this venue that shaped research clusters.

Foundational

Influential

Related Venues

Venues with the most similar research cluster distributions.

Papers

TitleYear
Shared Parameter Subspaces and Cross-Task Linearity in Emergently Misaligned Behaviour2025
Signal in the Noise: Polysemantic Interference Transfers and Predicts Cross-Model Influence2025
Small Vectors, Big Effects: A Mechanistic Study of RL-Induced Reasoning via Steering Vectors2025
Some Attention is All You Need for Retrieval2025
Sparse Autoencoders Trained on the Same Data Learn Different Features2025
Sparse but Wrong: Incorrect L0 Leads to Incorrect Features in Sparse Autoencoders2025
Spectral Dynamics in Neural Network Training: Mathematical Foundations for Understanding Representational Development2025
Steering Evaluation-Aware Language Models to Act Like They Are Deployed2025
Steering Out-of-Distribution Generalization with Concept Ablation Fine-Tuning2025
Stream: Scaling up Mechanistic Interpretability to Long Context in LLMs via Sparse Attention2025
SuperActivators: Only the Tail of the Distribution Contains Reliable Concept Signals2025
The Anatomy of Alignment: Decomposing Preference Optimization by Steering Sparse Features2025
The Dual Power of Interpretable Token Embeddings: Jailbreaking Attacks and Defenses for Diffusion Model Unlearning2025
The Geometry of Self-Verification in a Task-Specific Reasoning Model2025
The Impossibility of Inverse Permutation Learning in Transformer Models2025
Thinking Hard, Going Misaligned: Emergent Misalignment in LLMs2025
Thought Anchors: Which LLM Reasoning Steps Matter?2025
Thought Branches: Interpreting LLM Reasoning Requires Resampling2025
Three Desiderata for Faithfulness in Machine Learning Explanations: The Case for Causal Abstraction2025
Token Entanglement in Subliminal Learning2025
TopKLoRA2025
Towards a Mechanistic Understanding of Robustness in Finetuned Reasoning Models2025
Towards Understanding Multimodal Fine-Tuning: A Case Study into Spatial Features2025
Towards Understanding the Nature of Attention with Low-Rank Sparse Decomposition2025
Training Reliable Activation Probes With a Handful of Positive Examples2025
Uncovering Object Localization Mechanisms in VLMs2025
Understanding sparse autoencoder scaling in the presence of feature manifolds2025
Universal Neurons in GPT-2: Emergence, Persistence, and Functional Impact2025
Unsupervised decoding of encoded reasoning using language model interpretability2025
Vector Arithmetic in Concept and Token Subspaces2025
Visual Exploration of Feature Relationships in Sparse Autoencoders with Curated Concepts2025
What Affects the Effective Depth of Large Language Models?2025
What Do Refusal Tokens Learn? Fine-Grained Representations and Evidence for Downstream Steering2025
When seeing Overrides Knowing: Disentangling Knowledge Conflicts in Vision-Language Models2025
Where's the Bug? Attention Probing for Scalable Fault Localization2025
Who is In Charge? Dissecting Role Conflicts in LLM Instruction Following2025

« prev Page 3/3