In 160M and 290M parameter models, a new residual-stream split into scratch and protected channels causes massive activations to re-emerge in the protected decode channel, more concentrated on the start token.
All bark and no bite: Rogue dimensions in transformer language models obscure representational quality
7 Pith papers cite this work. Polarity classification is still indexing.
7
Pith papers citing it
citation-role summary
background 2
citation-polarity summary
roles
background 2representative citing papers
Cosine-scored SAEs with a learned direction-magnitude blend learn more concept-aligned features than standard inner-product SAEs at matched reconstruction quality.
Massive activations first appear in a single ME Layer due to RMSNorm and FFN, remain invariant thereafter, and a simple softening method raises LLM performance while reducing attention sinks.
Massive activations are constant large values in LLMs that function as indispensable bias terms and concentrate attention probabilities on specific tokens.
Training per-layer affine probes on frozen transformers yields more reliable latent predictions than the logit lens and enables detection of malicious inputs from prediction trajectories.
LLM.int8() performs 8-bit inference for transformers up to 175B parameters with no accuracy loss by combining vector-wise quantization for most features with 16-bit mixed-precision handling of systematic outlier dimensions.
HyperLens reveals that deeper transformer layers magnify small confidence changes into fine-grained trajectories, allowing quantification of cognitive effort where complex tasks demand more and standard SFT can reduce it.
citing papers explorer
-
Massive Activations Are Architecturally Robust: A Controlled Scratch/Commitment Residual Stream Test
In 160M and 290M parameter models, a new residual-stream split into scratch and protected channels causes massive activations to re-emerge in the protected decode channel, more concentrated on the start token.
-
Size Doesn't Matter: Cosine-Scored Sparse Autoencoders
Cosine-scored SAEs with a learned direction-magnitude blend learn more concept-aligned features than standard inner-product SAEs at matched reconstruction quality.
-
A Single Layer to Explain Them All:Understanding Massive Activations in Large Language Models
Massive activations first appear in a single ME Layer due to RMSNorm and FFN, remain invariant thereafter, and a simple softening method raises LLM performance while reducing attention sinks.
-
Massive Activations in Large Language Models
Massive activations are constant large values in LLMs that function as indispensable bias terms and concentrate attention probabilities on specific tokens.
-
Eliciting Latent Predictions from Transformers with the Tuned Lens
Training per-layer affine probes on frozen transformers yields more reliable latent predictions than the logit lens and enables detection of malicious inputs from prediction trajectories.
-
LLM.int8(): 8-bit Matrix Multiplication for Transformers at Scale
LLM.int8() performs 8-bit inference for transformers up to 175B parameters with no accuracy loss by combining vector-wise quantization for most features with 16-bit mixed-precision handling of systematic outlier dimensions.
-
HyperLens: Quantifying Cognitive Effort in LLMs with Fine-grained Confidence Trajectory
HyperLens reveals that deeper transformer layers magnify small confidence changes into fine-grained trajectories, allowing quantification of cognitive effort where complex tasks demand more and standard SFT can reduce it.