REVIEW 18 cited by
Uncovering mesa-optimization algorithms in Transformers
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
read the original abstract
Some autoregressive models exhibit in-context learning capabilities: being able to learn as an input sequence is processed, without undergoing any parameter changes, and without being explicitly trained to do so. The origins of this phenomenon are still poorly understood. Here we analyze a series of Transformer models trained to perform synthetic sequence prediction tasks, and discover that standard next-token prediction error minimization gives rise to a subsidiary learning algorithm that adjusts the model as new inputs are revealed. We show that this process corresponds to gradient-based optimization of a principled objective function, which leads to strong generalization performance on unseen sequences. Our findings explain in-context learning as a product of autoregressive loss minimization and inform the design of new optimization-based Transformer layers.
Forward citations
Cited by 18 Pith papers
-
Training with (Swap) Regret Loss in a Single-Layer Self-Attention Model: A Case Study on the Probability Simplex
Training single-layer attention with squared regret loss has stationary points that implement smoothed fictitious play (external regret) and, via a new swap-regret loss, the Blum–Mansour no-swap-regret algorithm.
-
Induction Heads Interpolate N-Grams
Induction-head circuits implement soft context-matching (Jelinek–Mercer-style interpolation over partial matches) plus BOS-induced Dirichlet pseudo-counts, and trained transformers recover both mechanisms.
-
The Orthogonalized Read Is a Removable Training Scaffold for Recurrent Memory
The orthogonalized read on noisy recall is a removable training scaffold: it multiplies escape hazard ~6.4x during plateau training, but a stock mLSTM achieves full accuracy once the read is annealed away.
-
Phase Transitions in Attention: A Bayesian Theory of Copy Head Emergence
Bayesian reduction of attention posterior on copy task predicts first-order phase transition for softmax attention and second-order followed by crossover for linear attention.
-
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.
-
Beyond Test-Time Memory: State-Space Optimal Control for LLM Reasoning
A symplectic LQR layer inserted as an adapter into pretrained LLMs yields large gains on MATH-500, AMC and AIME by solving a latent optimal-control problem at inference time.
-
Sycophancy to Subterfuge: Investigating Reward-Tampering in Large Language Models
LLMs trained on simple specification gaming generalize to zero-shot reward tampering including rewriting their own reward function.
-
Grounding latent algorithm routing in transformer reasoning
On ROUTEBENCH, 306M–612M dense transformers close most of the oracle routing gap with decodable, nuisance-stable, intervention-sensitive solver-family preferences.
-
Dynamic Short Convolutions Improve Transformers
Dynamic short convolutions applied to key/query/value projections and linear layers in Transformers yield consistent performance gains and 1.33-1.60x compute advantages over standard models on language modeling from 1...
-
Blurry Window Attention
Blurry Window Attention stores a frequency window and reconstructs blurry KV history via Dirichlet kernel interpolation, achieving 8x better state efficiency than sliding window attention on the MQAR synthetic task.
-
A Systematic Study of Behavioral Cloning for Scientific Data Annotation
Introduces 9 synthetic annotation tasks and benchmarks for behavioral cloning, finding hierarchical skill learning, scaling benefits, effective multi-task pretraining, and shared internal representations of task phase...
-
OSDN: Improving Delta Rule with Provable Online Preconditioning in Linear Attention
OSDN adds online diagonal preconditioning to the Delta Rule, preserving chunkwise parallelism while proving super-geometric convergence and delivering 32-39% recall gains at 340M-1.3B scales.
-
Correcting Influence: Unboxing LLM Outputs with Orthogonal Latent Spaces
A latent mediation framework with sparse autoencoders enables non-additive token-level influence attribution in LLMs by learning orthogonal features and back-propagating attributions.
-
Learning to Remember, Learn, and Forget in Attention-Based Models
Palimpsa adds a per-slot importance/precision state to gated linear attention, letting a fixed-size memory forget stale information and protect important information, and recovers Mamba2 as a high-forgetting limit.
-
Selective Induction Heads: How Transformers Select Causal Structures In Context
Transformers can learn to select the correct lag of an interleaved Markov chain in context via a circuit the authors call a selective induction head, whose asymptotic optimality proof is incomplete.
-
MDN: Parallelizing Stepwise Momentum for Delta Linear Attention
MDN parallelizes stepwise momentum for delta linear attention using geometric reordering and dynamical systems analysis, yielding performance gains over Mamba2 and GDN on 400M and 1.3B models.
-
Position: We Need An Algorithmic Understanding of Generative AI
The paper argues for a systematic algorithmic understanding of LLMs and presents a case study suggesting that Llama models do not implement BFS or DFS on graph navigation tasks.
-
Understanding Large Language Models
The paper reviews Transformer architecture, emergent LLM capabilities resembling cognition, explainable AI methods, and argues against both anthropomorphism and overly reductive views of LLM behavior as mere memorization.
Discussion (0). Sign in to comment.