REVIEW 23 cited by
Linear Transformers Are Secretly Fast Weight Programmers
Not yet reviewed by Pith; the record is open.
This paper has not been read by Pith yet. Machine review is queued; the pith claim, tier, and objections will appear here once it completes.
SPECIMEN: schema-true, not a live event
T0 review · schema-true
One-sentence machine reading of the paper's core claim.
pith:XXXXXXXX · record.json · timestamp
Linear Transformers Are Secretly Fast Weight Programmers
read the original abstract
We show the formal equivalence of linearised self-attention mechanisms and fast weight controllers from the early '90s, where a ``slow" neural net learns by gradient descent to program the ``fast weights" of another net through sequences of elementary programming instructions which are additive outer products of self-invented activation patterns (today called keys and values). Such Fast Weight Programmers (FWPs) learn to manipulate the contents of a finite memory and dynamically interact with it. We infer a memory capacity limitation of recent linearised softmax attention variants, and replace the purely additive outer products by a delta rule-like programming instruction, such that the FWP can more easily learn to correct the current mapping from keys to values. The FWP also learns to compute dynamically changing learning rates. We also propose a new kernel function to linearise attention which balances simplicity and effectiveness. We conduct experiments on synthetic retrieval problems as well as standard machine translation and language modelling tasks which demonstrate the benefits of our methods.
Forward citations
Cited by 23 Pith papers
-
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.
-
WriteSAE: Sparse Autoencoders for Recurrent State
WriteSAE decomposes recurrent model cache writes into substitutable atoms with a closed-form logit shift, achieving high substitution success and targeted behavioral installs on models like Qwen3.5 and Mamba-2.
-
WriteSAE: Sparse Autoencoders for Recurrent State
WriteSAE is the first sparse autoencoder that factors decoder atoms into the native d_k x d_v cache write shape of recurrent models and supplies a closed-form per-token logit shift for atom substitution.
-
Verbalizable Representations Form a Global Workspace in Language Models
Language models represent their current reasoning in a small, readable set of verbalizable vectors (the J-space) that functions like a global workspace.
-
Tensor Cache: Eviction-conditioned Associative Memory for Transformers
Tensor Cache augments sliding-window attention with an eviction-fed outer-product associative memory and a training correction to improve long-context performance under bounded memory.
-
WriteSAE: Sparse Autoencoders for Recurrent State
WriteSAE factors sparse autoencoder decoder atoms to the native d_k x d_v cache write shape in recurrent models, provides a closed-form logit shift, and demonstrates high success in atom substitution and behavioral ed...
-
Learning, Fast and Slow: Towards LLMs That Adapt Continually
Fast-Slow Training uses context optimization as fast weights alongside parameter updates as slow weights to achieve up to 3x better sample efficiency, higher performance, and less catastrophic forgetting than standard...
-
Preconditioned DeltaNet: Curvature-aware Sequence Modeling for Linear Recurrences
Preconditioned delta-rule models with a diagonal curvature approximation improve upon standard DeltaNet, GDN, and KDA by better approximating the test-time regression objective.
-
Infrared Organization and Critical Cognitive Field Formation in Transformer Dynamics
Pythia layer Jacobians reorganize during training into an approximately flat infrared TDOS ρ(λ)∼λ^{-0.1} with K(t)∼1/t kernels and a transient memory-self-energy maximum interpreted as critical cognitive-field formation.
-
Sparse Delta Memory: Scaling the State of Linear RNNs through Sparsity
SDM sparsifies the Gated DeltaNet update rule to enable 1000x larger recurrent memory states at iso-FLOP, improving long-context recall and short-context reasoning over GDN and matching full attention at 8B scale.
-
A Hippocampus for Linear Attention: An Exact Memory for What the Recurrent State Forgets
HOLA pairs a compressive delta-rule recurrent state with a residual-selected exact KV cache and decoupled RMSNorm-gamma read, yielding lower perplexity than both standard linear attention and full-attention baselines ...
-
Test-Time Training with Next-Token Prediction
TTT-NTP adapts pretrained LLMs at test time by training fast weights to match next-position hidden states from the forward pass, yielding consistent gains on long-context benchmarks across Llama, Mistral, and Qwen models.
-
Memory by Design: Probabilistic Sequence Layers
The design-model framework unifies sub-quadratic sequence models as Bayesian filters and introduces a covariance-tracking Bayesian Layer that improves retrieval robustness beyond training regimes on MQAR and RULER benchmarks.
-
Multi-Mixer Models: Flexible Sequence Modeling with Shared Representations
Oryx hybridizes attention and linear recurrent mixers along the sequence axis with high parameter sharing, outperforming single-mixer baselines on language modeling and retrieval at up to 1.4B scale under mixed training.
-
Learning, Fast and Slow: Towards LLMs That Adapt Continually
Fast-Slow Training combines slow parameter updates with fast context optimization to achieve up to 3x better sample efficiency, higher performance, less forgetting, and preserved plasticity in continual LLM learning.
-
A Single-Layer Model Can Do Language Modeling
A 130M-parameter 1-layer GPN achieves FineWeb-Edu perplexity 18.06, within 13% of a 12-layer Transformer++ (16.05) and 18% of a 10-layer GDN (15.34).
-
Memoir: Should a Model Write to Its Memory While It Thinks?
Writing to fast memory during pondering slows associative-recall learning at a fixed budget, but does not reduce final performance once training is long enough.
-
StateX: Enhancing RNN Recall via Post-training State Expansion
StateX post-trains RNNs to expand recurrent state size, improving recall and in-context learning with negligible parameter growth.
-
Infrared Organization and Critical Cognitive Field Formation in Transformer Dynamics
Slow relaxation modes in Pythia transformers accumulate toward zero rate during training, yielding a near-flat infrared spectrum and 1/t memory kernels—but the claimed 'critical cognitive field formation' is not direc...
-
Reversible Foundations: Training a 120B Sparse MoE through State-Preserving Scaling
A 120B sparse MoE model with 460 experts was trained on one 8-GPU node to loss 1.78 using reversible recurrence and state-preserving scaling from a 1.78B dense seed, with 5.93B active parameters.
-
Cognitive Field Theory: Memory-Dressed Collective Dynamics of Intelligence
The paper asserts that Hopfield networks, RNNs, transformers, and the author's FHRN model are all special cases of a single stochastic field equation whose collective time-scale spectrum governs cognition.
-
Cognitive Field Theory: Memory-Dressed Collective Dynamics of Intelligence
A field-theoretic reframing in which cognition is the spectrum of slow collective modes of generic Langevin dynamics, whose central one-loop result is assumed and whose headline memory-dressing claims do not appear in...
-
A Vision Toward Energy-Efficient Domain-Specific Artificial Intelligence Models and Agents
A position paper proposing compact, domain-specific AI agents as the path to ≥1000× energy efficiency, without demonstrating the claim.
discussion (0)
Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.