Slaying the Hydra: Interaction-Aware Circuit Discovery in Language Models
cs.LG, cs.AI, cs.CL, stat.ML
Submitted: 2026-10-02
Updated: 2026-10-02
Sources
- Matryoshka attribution: Learning to attribute language model outputs to representations and weights
- Conditional Co-Ablation: Recovering Self-Repair Backups in Transformer Circuits
- How to use and interpret activation patching
- MIB: A Mechanistic Interpretability Benchmark
- Fine-Tuning Enhances Existing Mechanisms: A Case Study on Entity Tracking
- A Computationally Feasible Framework for Causal Probabilistic Explanation
- The Curse of Multiple Mediators: Hidden Interaction Effects in Activation Patching
Related papers
- Polynomial-Augmented Neural Networks (PANNs) with Weak Orthogonality Constraints for Enhanced Function and PDE Approximation
- AIRL-S: Unifying Reinforcement Learning and Search-Based Test-Time Scaling via Adversarial Inverse Reinforcement Learning
- Transformers as Bayesian In-Context Experimenters: Smoothness-Adaptive Efficient ATE Estimation
- Convergence issues in Relational Concept Analysis based on AOC-posets
- Beliefs Beyond Posteriors: The Predictive Cost of ELBO Binding
- Understanding Diffusion Models via Ratio-Based Function Approximation with SignReLU Networks