Re-derivation of activation patching NIE reveals it captures interaction effects in addition to direct causal effects, demonstrated via GPT-2 IOI circuit where INT explains component ranking issues and faithfulness instability.
SAEs Are Good for Steering – If You Select the Right Features , url=
6 Pith papers cite this work. Polarity classification is still indexing.
6
Pith papers citing it
years
2026 6verdicts
UNVERDICTED 6representative citing papers
Probes predicting future behaviors from intermediate steps enable Future Probe Controlled Generation for steering large reasoning models with minimal quality degradation.
Query Lens extends Logit Lens to interpret sparse features via key-value analysis and indirect effects, yielding coherent token signatures where Logit Lens fails, and proposes the Subspace Channel Hypothesis.
Introduces BMC, a manifold bandit framework that organizes problems into a hierarchical task tree and applies Bayesian learning to balance productivity, diversity, and utility in LLM curriculum sampling.
Qwen-Scope provides open-source sparse autoencoders for Qwen models that function as practical interfaces for steering, evaluating, data workflows, and optimizing large language models.
Trains and releases SAEs for Qwen3-1.7B/4B/8B models with layer-wise coverage and demonstrates causal steering of refusal via selected features.
citing papers explorer
-
The Curse of Multiple Mediators: Hidden Interaction Effects in Activation Patching
Re-derivation of activation patching NIE reveals it captures interaction effects in addition to direct causal effects, demonstrated via GPT-2 IOI circuit where INT explains component ranking issues and faithfulness instability.
-
Predicting Future Behaviors in Reasoning Models Enables Better Steering
Probes predicting future behaviors from intermediate steps enable Future Probe Controlled Generation for steering large reasoning models with minimal quality degradation.
-
Query Lens: Interpreting Sparse Key-Value Features with Indirect Effects
Query Lens extends Logit Lens to interpret sparse features via key-value analysis and indirect effects, yielding coherent token signatures where Logit Lens fails, and proposes the Subspace Channel Hypothesis.
-
Manifold Bandits: Bayesian Curriculum Learning over the Latent Geometry of Large Language Models
Introduces BMC, a manifold bandit framework that organizes problems into a hierarchical task tree and applies Bayesian learning to balance productivity, diversity, and utility in LLM curriculum sampling.
-
Qwen-Scope: Turning Sparse Features into Development Tools for Large Language Models
Qwen-Scope provides open-source sparse autoencoders for Qwen models that function as practical interfaces for steering, evaluating, data workflows, and optimizing large language models.
-
Discovering Millions of Interpretable Features with Sparse Autoencoders
Trains and releases SAEs for Qwen3-1.7B/4B/8B models with layer-wise coverage and demonstrates causal steering of refusal via selected features.