The Key to Going Linear: Analysis-Driven Transformer Linearization

• URL: http://arxiv.org/abs/2607.07706v1
• Date: Wed, 08 Jul 2026 17:59:09 GMT
• Title: The Key to Going Linear: Analysis-Driven Transformer Linearization
• Abstract summary: We show that softmax relies on key-dependent, rank-1 projections, elucidating why delta-style networks outperform purely gated accumulation.<n>We identify a potential source of approximation errors and introduce structural interventions, specifically sink tokens, short convolutions, and fixed-budget cache routing.
• Score: 8.56204304015324
• License: http://creativecommons.org/licenses/by-nc-nd/4.0/
• Abstract: The quadratic cost of causal self-attention severely bottlenecks long-context transformer inference. While numerous post hoc linearization pipelines exist, it is difficult to identify which components preserve model quality. This work isolates the effect of state update design in a strict frozen-backbone regime. We show that softmax relies on key-dependent, rank-1 orthogonal projections, elucidating why delta-style networks outperform purely gated accumulation. We identify a potential source of approximation errors and introduce structural interventions, specifically sink tokens, short convolutions, and fixed-budget cache routing, which reduces the remaining gap. We scale this linearization approach across LLaMA and Qwen models up to 32B parameters, outperforming prior post hoc baselines on MMLU and matching the long-context retrieval of complex adaptive-caching frameworks.

Related papers

• CRISP: Cliff-awaRe Input-adaptive Sparse Prefilling with Structural-Mass-Motivated Routing [90.63350360476294]
  We present CRISP (Cliff-awaRe Input-adaptive Sparse Prefilling), which identifies and addresses two structural challenges in this dynamic routing paradigm.<n>We show that the routing decision can be read directly off the structure of the proxy attention map.<n>We formalize the post-softmax mass cliff and demonstrate theoretically that strictly cumulative coverage thresholds accumulate O(n) background noise at long contexts.
  arXiv  Detail & Related papers  (2026-09-01T22:49:46Z)
• Scaling Interpretable Transformers with Parity Bottleneck Layers [0.0]
  We introduce a GPT-2-scale architecture whose intermediate representations are efficient and wide / sparse by design.<n>A Deep Parity Bottleneck (DPB) replaces a learned over-complete basis with a parameter-free algebraic dictionary.<n>The ParityTransformer's features are native to the model's forwards pass by construction, addressing the question of whether SAEs probe features the model actually uses during computation.
  arXiv  Detail & Related papers  (2026-07-22T18:25:16Z)
• Closed-Form Spectral Regularization for Multi-Task Model Merging [96.82449201305234]
  Model merging combines several independently fine-tuned experts into a single multi-task model without any training data.<n>State-of-the-art merging methods formulate merging as a layer-wise interference problem.<n>We propose SWUDI, a closed-form method that combines a soft exponential filter, which matches the gradient-flow trajectory of iterative descent.
  arXiv  Detail & Related papers  (2026-06-05T14:00:47Z)
• Clipping Bottleneck: Stabilizing RLVR via Stochastic Recovery of Near-Boundary Signals [83.0127582612634]
  Near-boundary Rescue (NSR) is a minimal, plug-and-play modification that retains out-of-bound tokens to recover lost signals.<n>NSR substantially improves training stability and delivers consistent gains over strong baselines such as DAPO and GSPO.
  arXiv  Detail & Related papers  (2026-05-21T16:45:31Z)
• COREY: Entropy-Guided Runtime Chunk Scheduling for Selective Scan Kernels [11.316541559874864]
  Prototype scheduler uses activation entropy estimated via fixed-width histograms as a runtime signal for chunk-size selection.<n>COREY is positioned as a Concept and Feasibility contribution.<n>This work does not yet include a fully integrated end-to-end run connecting Tier 2a and Tier 2b.
  arXiv  Detail & Related papers  (2026-04-12T12:07:48Z)
• Graph-RHO: Critical-path-aware Heterogeneous Graph Network for Long-Horizon Flexible Job-Shop Scheduling [26.97705087364345]
  Graph-RHO is a novel critical-path-aware graph-based RHO framework.<n>We introduce a topology-aware heterogeneous graph network that encodes subproblems as operation-machine graphs with multi-relational edges.<n>We incorporate a critical-path-aware mechanism that injects inductive biases during training to distinguish highly sensitive bottleneck operations from robust ones.
  arXiv  Detail & Related papers  (2026-04-11T07:35:09Z)
• Memory-Conditioned Flow-Matching for Stable Autoregressive PDE Rollouts [0.0]
  Autoregressive generative PDE solvers can be accurate one step ahead yet drift over long rollouts.<n>We show that eliminating unresolved variables yields an exact resolved evolution with a Markov term.<n>We then derive discrete Grnwall rollout bounds that separate memory approximation from conditional generation error.
  arXiv  Detail & Related papers  (2026-02-06T13:21:52Z)
• READER: Retrieval-Assisted Drafter for Efficient LLM Inference [0.0386965802948046]
  Autoregressive Language Models instantiate a factorized likelihood over token sequences, yet their strictly sequential decoding process imposes an intrinsic lower bound on latency inference.<n>This bottleneck has emerged as a central obstacle to the scalable deployment of large-scale generative models.<n>We present READER, a speculative decoding framework that bypasses the training of the auxiliary draft model.
  arXiv  Detail & Related papers  (2025-08-12T16:47:48Z)
• Beyond Homogeneous Attention: Memory-Efficient LLMs via Fourier-Approximated KV Cache [67.47789629197857]
  We propose a training-free framework that exploits the heterogeneous roles of transformer head dimensions.<n>By projecting the long-context-insensitive dimensions onto Fourier bases, FourierAttention approximates their temporal evolution with fixed-length spectral coefficients.<n>We show that FourierAttention achieves the best long-context accuracy on LongBench and Needle-In-A-Haystack.
  arXiv  Detail & Related papers  (2025-06-13T15:35:54Z)
• Gated Attention for Large Language Models: Non-linearity, Sparsity, and Attention-Sink-Free [81.65559031466452]
  We conduct experiments to investigate gating-augmented softmax attention variants.<n>We find that a simple modification-applying a head-specific sigmoid gate after the Scaled Dot-Product Attention (SDPA)-consistently improves performance.
  arXiv  Detail & Related papers  (2025-05-10T17:15:49Z)
• ThinK: Thinner Key Cache by Query-Driven Pruning [63.13363917871414]
  Large Language Models (LLMs) have revolutionized the field of natural language processing, achieving unprecedented performance across a variety of applications.<n>This paper focuses on the long-context scenario, addressing the inefficiencies in KV cache memory consumption during inference.<n>We propose ThinK, a novel query-dependent KV cache pruning method designed to minimize attention weight loss while selectively pruning the least significant channels.
  arXiv  Detail & Related papers  (2024-07-30T17:59:08Z)
• Haar Wavelet based Block Autoregressive Flows for Trajectories [129.37479472754083]
  Prediction of trajectories such as that of pedestrians is crucial to the performance of autonomous agents. We introduce a novel Haar wavelet based block autoregressive model leveraging split couplings. We illustrate the advantages of our approach for generating diverse and accurate trajectories on two real-world datasets.
  arXiv  Detail & Related papers  (2020-09-21T13:57:10Z)
• Generalizing Variational Autoencoders with Hierarchical Empirical Bayes [6.273154057349038]
  We present Hierarchical Empirical Bayes Autoencoder (HEBAE), a computationally stable framework for probabilistic generative models. Our key contributions are two-fold. First, we make gains by placing a hierarchical prior over the encoding distribution, enabling us to adaptively balance the trade-off between minimizing the reconstruction loss function and avoiding over-regularization.
  arXiv  Detail & Related papers  (2020-07-20T18:18:39Z)

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.