WriteSAE introduces sparse autoencoders with rank-1 matrix atoms for recurrent state updates, allowing replacement tests that outperform deletion on 92.4% of positions and a formula predicting logit changes with R²=0.98.
hub Mixed citations
org/abs/2305.01610
Mixed citation behavior. Most common role is background (60%).
25
Pith papers citing it
Background
60% of classified citations
hub tools
citation-role summary
background 3
method 2
citation-polarity summary
representative citing papers
AVLLMs encode audio semantics in middle layers but suppress them in final text outputs when audio conflicts with vision, due to training that largely inherits from vision-language base models.
Neurons exhibit concept-conditioned activation ranges forming Gaussian-like distributions with minimal overlap, and range-based interventions via NeuronLens outperform neuron-level masking in targeted manipulation with reduced collateral effects.
K-sparse autoencoders with dead-latent fixes produce clean scaling laws and better feature quality metrics that improve with size, shown by training a 16-million-latent model on GPT-4 activations.
LLMs recover dominant binomial orders from corpora but align less closely with exact preference distributions, with preference strength partially encoded in middle-to-late layers and manipulable via steering.
Critical percolation clusters embedded in high dimensions, combined with taxonomic latent variables, form an analytically tractable synthetic data model whose ground-truth hierarchy can be linearly decoded from network activations.
LLM representations encode essay quality in a linearly decodable form that emerges across layers and includes identifiable scoring neurons whose distribution shifts with essay length.
ICALens applies an optimized ICA workflow to LLM activations and recovers compact interpretable directions that match or exceed public SAEs on SAEBench probing and perturbation tasks without per-layer dictionary training.
A linear probe trained on 190k congressional tweets identifies a partisan direction in Llama 3.1 8B layer 18 that can be causally ablated or amplified to reverse or shift the model's political output.
Formalizes concept learning in sparse autoencoders as set alignment between human-defined and model-induced concepts, distinguishing detection, separation, and approximation with geometric conditions for neuron representation.
Activation steering produces synthetic safety-violating data that improves downstream classifiers over prompting on most tested concepts when a harmonic mean of alignment, coherence, and diversity is optimized.
A vision transformer for runway keypoint regression is decomposed via K-SVD into content and style atoms; the model relies primarily on content atoms, enabling out-of-model-scope detection for runtime assurance in aviation.
Chessformer is a unified encoder-only transformer for chess that uses square tokens, geometric attention bias, and an attention-based policy head to set new records in human move prediction accuracy, playing strength, and interpretability.
Suppressing one refusal neuron or amplifying one concept neuron bypasses safety alignment in LLMs from 1.7B to 70B parameters without training or prompt engineering.
Tree SAE learns hierarchical feature structures by combining activation coverage with a new reconstruction condition, outperforming prior SAEs on hierarchical pair detection while matching state-of-the-art benchmark performance.
Perturbation probing identifies tiny sets of FFN neurons that control refusal templates and language routing in LLMs, enabling precise ablations and directional interventions that alter behavior on benchmarks while preserving safety.
LIRA aligns latent instruction representations in LLMs to defend against jailbreaks, backdoors, and undesired knowledge, blocking over 99% of PEZ attacks and achieving optimal WMDP forgetting.
Pretrained vision transformers use specific attention heads sensitive to Gestalt continuity for object binding, shown via probes on synthetic datasets and ablation experiments.
GPT-2 Small's terminal MLP implements a legible three-tier exception handler with 27 named neurons that routes predictions, while previously identified knowledge neurons function as amplifiers of residual-stream signals rather than fact storage.
Language models contain localized entity-selective neurons in early layers that causally mediate factual recall for specific entities across surface variations.
MIST models up to 10x larger than prior work, fine-tuned on over 400 structure-property tasks, match or exceed SOTA on benchmarks and demonstrate zero-shot olfactory perception mapping consistent with hyperbolic geometry.
Introduces Modality Dominance Score (MDS) to measure modality-specific features in VLMs and applies training-free editing to improve bias mitigation, adversarial generation, and modality control.
Safe-SAIL supplies a pre-explanation metric and segment-level simulation to interpret 1758 safety SAE features across pornography, politics, violence, and terror, with public models and tools released.
Varying evaluation metrics and corruption methods in activation patching produces different localization and circuit discovery outcomes in language models, leading to recommendations for preferred practices.
citing papers explorer
-
WriteSAE: Sparse Autoencoders for Recurrent State
WriteSAE introduces sparse autoencoders with rank-1 matrix atoms for recurrent state updates, allowing replacement tests that outperform deletion on 92.4% of positions and a formula predicting logit changes with R²=0.98.
-
A Single Neuron Is Sufficient to Bypass Safety Alignment in Large Language Models
Suppressing one refusal neuron or amplifying one concept neuron bypasses safety alignment in LLMs from 1.7B to 70B parameters without training or prompt engineering.