ReSAE: Residualized Sparse Autoencoders for Multi-Layer Transformer Interventions

• URL: http://arxiv.org/abs/2605.27819v1
• Date: Wed, 27 May 2026 01:27:53 GMT
• Title: ReSAE: Residualized Sparse Autoencoders for Multi-Layer Transformer Interventions
• Abstract summary: We introduce Residualized Sparse Autoencoders (ReSAEs)<n>ReSAEs fit an affine map between selected layers and train each later-layer SAE on the unexplained residual rather than on the full activation.<n>On Pythia-1.4B and Gemma-2-9B, residualization reduces decoder redundancy and improves sparse probing and targeted perturbations.
• Score: 10.217703037524824
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: Sparse autoencoders are usually trained one layer at a time, even though transformer residual stream activations are strongly coupled across depth. This creates a practical problem for multi-layer interventions: different layerwise dictionaries can spend capacity representing the same carried-forward information, and replacing several layers at once can produce interactions that are not predicted by single-layer behavior. We introduce Residualized Sparse Autoencoders (ReSAEs), which fit an affine map between selected layers and train each later-layer SAE on the unexplained residual rather than on the full activation. Reconstructions are mapped back into the original activation space through the fitted affine chain, so ReSAEs can be evaluated with the same intervention protocols as ordinary SAEs. On Pythia-1.4B and Gemma-2-9B, residualization reduces decoder redundancy and improves sparse probing and targeted perturbation in most tested settings. Despite reconstructing less of the raw activation variance, ReSAEs recover more transformer cross entropy under multi-layer replacement. This gain is clearest under teacher-forcing and at sufficient sparsity online, indicating that ReSAEs preserve the components of the activation most relevant to the model's downstream computation. These results suggest that removing linearly predictable cross-layer structure is a useful default for multi-layer SAE interventions.

Related papers

• Decoder-Preserving Sparse Autoencoders: Which Readouts Survive Sparse Compression? [0.0]
  We formalize this ambiguity as a matrix-valued distortion between optimal ridge-prediction operators and train decoder-preserving SAEs.<n>A controlled sparse experiment shows that a declared prior protects held-out combinations from its task subspace.<n>These results show that reconstruction quality does not determine which refitted linear readouts survive sparse compression.
  arXiv  Detail & Related papers  (2026-07-19T22:19:17Z)
• T^2MLR: Transformer with Temporal Middle-Layer Recurrence [28.818755729513793]
  We develop a transformers-based latent reasoning architecture that fuses a cached middle layer representation from the previous token directly into an earlier layer of the current token position.<n>Applying recurrence to only a localized middle-layer block (as little as 20% of the network) often outperforms full-layer recurrence.<n>These results suggest that effective latent reasoning in Transformers does not require looping over all layers as in previous works, but can instead emerge more strongly from targeted middle-layer recurrence.
  arXiv  Detail & Related papers  (2026-07-16T16:33:59Z)
• Latent Recurrent Transformer: Architecture Exploration, Training Strategies, and Scaling Behavior [107.2098567818173]
  Latent Recurrent Transformer (LRT) is a lightweight augmentation of autoregressive transformers.<n>LRT reuses a high-level source-layer hidden state from the previous token as recurrent memory for the next token.
  arXiv  Detail & Related papers  (2026-05-26T10:10:26Z)
• Replacement Learning: Training Neural Networks with Fewer Parameters [15.09968642484538]
  Replacement Learning (RepL) is a training-time paradigm that reduces full-depth redundancy by replacing selected blocks.<n>RepL preserves local contextual continuity while avoiding unnecessary full-layer.<n>We show that RepL reduces trainable parameters, GPU memory usage, and training time while matching or surpassing standard end-to-end training.
  arXiv  Detail & Related papers  (2026-05-19T08:34:31Z)
• scHelix: Asymmetric Dual-Stream Integration via Explicit Gene-Level Disentanglement [84.88065404629079]
  scHelix is a dataset-adaptive framework that fundamentally changes how features are processed.<n> scHelix utilizes a dual-stream sparse diffusion encoder equipped with stop-gradient graph caching.<n> scHelix outperforms state-of-the-art methods.
  arXiv  Detail & Related papers  (2026-05-18T15:55:13Z)
• Attribution-Guided Distillation of Matryoshka Sparse Autoencoders [0.0]
  We introduce Distilled Matryoshka Sencoders (DMSAEs), a training pipeline that distills a compact core of consistently useful features and reuses it to train new SAEs.<n>DMSAEs run an iterative distillation cycle: train a Matryoshka SAE with a shared core, use gradient X activation to measure each feature's contribution to next-token loss in the most nested reconstruction, and keep only the smallest subset that explains a fixed fraction of the attribution.<n>On Gemma-2-2B layer 12 residual stream activations, seven cycles of distillation (500M tokens, 65k width)
  arXiv  Detail & Related papers  (2025-12-31T17:12:55Z)
• OmniSAT: Compact Action Token, Faster Auto Regression [70.70037017501357]
  We introduce an Omni Swift Action Tokenizer, which learns a compact, transferable action representation.<n>The resulting discrete tokenization shortens the training sequence by 6.8$times$, and lowers the target entropy.
  arXiv  Detail & Related papers  (2025-10-08T03:55:24Z)
• Regularizing Subspace Redundancy of Low-Rank Adaptation [54.473090597164834]
  We propose ReSoRA, a method that explicitly models redundancy between mapping subspaces and adaptively Regularizes Subspace redundancy of Low-Rank Adaptation.<n>Our proposed method consistently facilitates existing state-of-the-art PETL methods across various backbones and datasets in vision-language retrieval and standard visual classification benchmarks.<n>As a training supervision, ReSoRA can be seamlessly integrated into existing approaches in a plug-and-play manner, with no additional inference costs.
  arXiv  Detail & Related papers  (2025-07-28T11:52:56Z)
• Residual Stream Analysis with Multi-Layer SAEs [21.142967037533175]
  We introduce the multi-layer SAE (MLSAE), a single SAE trained on the residual stream activation vectors from every transformer layer.<n>We find that individual latents are often active at a single layer for a given token or prompt, but the layer at which an individual latent is active may differ for different tokens or prompts.<n>Our results represent a new approach to understanding how representations change as they flow through transformers.
  arXiv  Detail & Related papers  (2024-09-06T11:01:55Z)
• Recurrence Boosts Diversity! Revisiting Recurrent Latent Variable in Transformer-Based Variational AutoEncoder for Diverse Text Generation [85.5379146125199]
  Variational Auto-Encoder (VAE) has been widely adopted in text generation. We propose TRACE, a Transformer-based recurrent VAE structure.
  arXiv  Detail & Related papers  (2022-10-22T10:25:35Z)
• Deriving Differential Target Propagation from Iterating Approximate Inverses [91.3755431537592]
  We show that a particular form of target propagation, relying on learned inverses of each layer, which is differential, gives rise to an update rule which corresponds to an approximate Gauss-Newton gradient-based optimization. We consider several iterative calculations based on local auto-encoders at each layer in order to achieve more precise inversions for more accurate target propagation.
  arXiv  Detail & Related papers  (2020-07-29T22:34:45Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.