No article text was recorded for this version.
residual stream decomposition
Version 1 of 127 papers30 conceptsfrozen 2026-10-04
Papers in this version
- Blog: How to Dissect a Muppet: The Structure of Transformer Embedding Spaces
- Blog: Explaining How Transformers Use Context to Build Predictions
- Blog: Self-Disentanglement and Re-Composition for Cross-Domain Few-Shot Segmentation
- Blog: Evaluating Deep Taylor Decomposition for Reliability Assessment in the Wild
- Blog: Circuit Component Reuse Across Tasks in Transformer Language Models
- Blog: Register and [CLS] tokens induce a decoupling of local and global features in large ViTs
- Blog: Forbidden Facts: An Investigation of Competing Objectives in Llama 2
- Blog: Token-wise Decomposition of Autoregressive Language Model Hidden States for Analyzing Model Predictions
- Blog: A Closer Look at Backdoor Attacks on CLIP
- Blog: Why Bother with Geometry? On the Relevance of Linear Decompositions of Transformer Embeddings
- Blog: When Parts Are Greater Than Sums: Individual LLM Components Can Outperform Full Models
- Blog: Towards Opening the Black Box of Neural Machine Translation: Source and Target Interpretations of the Transformer
- Blog: Characterizing Mechanisms for Factual Recall in Language Models
- Blog: Interpreting CLIP's Image Representation via Text-Based Decomposition
- Blog: DePass: Unified Feature Attributing by Simple Decomposed Forward Pass
- Blog: Understanding Cross-layer Contributions to Mixture-of-Experts Routing in LLMs
- Blog: Measuring the Mixing of Contextual Information in the Transformer
- Blog: TuCo: Measuring the Contribution of Fine-Tuning to Individual Responses of LLMs
- Blog: When Do Transformers Learn Heuristics for Graph Connectivity?
- Blog: UniBias: Unveiling and Mitigating LLM Bias through Internal Attention and FFN Manipulation
- Blog: Detecting and Understanding Vulnerabilities in Language Models via Mechanistic Interpretability
- Blog: Incorporating Residual and Normalization Layers into Analysis of Masked Language Models
- Blog: Decomposing and Interpreting Image Representations via Text in ViTs Beyond CLIP
- Blog: Universal Redundancies in Time Series Foundation Models
- Blog: Locate then Correct: Debiasing Attention Heads in CLIP
- Blog: Decomposing LLM Computation with Jets
- Blog: DecompX: Explaining Transformers Decisions by Propagating Token Decomposition
History & citation
Current · Permanent link
- This direction split from Additive Decomposition of Transformer Predictions (89% overlap)
Cite this direction
Latest version (a living page):
@misc{lacuna_residual_stream_decomposition_b9ecaa18,
title = {residual stream decomposition},
author = {{Lacuna}},
howpublished = {Lacuna research direction},
year = {2026},
url = {https://lacuna.tiptreesystems.com/direction/residual-stream-decomposition/rdr_b9ecaa18be7f4ce0890e69b7c9c7445a},
note = {Latest version of a living research direction; cite a version for a fixed text. Accessed 2026-11-19},
}This exact version (version 1, frozen text):
@misc{lacuna_residual_stream_decomposition_b9ecaa18_v1,
title = {residual stream decomposition},
author = {{Lacuna}},
howpublished = {Lacuna research direction},
year = {2026},
url = {https://lacuna.tiptreesystems.com/direction/residual-stream-decomposition/rdr_b9ecaa18be7f4ce0890e69b7c9c7445a/v/1},
note = {Version 1, frozen 2026-10-04. Accessed 2026-11-19},
}