gradient-flow

Tag

Cards List
#gradient-flow

Faster Query-Key Learning Sharpens Attention in Self-Attention Models

arXiv cs.LG · 5d ago Cached

This paper analyzes how the parameterization of query-key and output-value circuits in self-attention models affects attention sharpness during training. Through gradient-flow analysis, they show that faster query-key learning relative to output-value learning produces sharper attention, improving interpretability without sacrificing predictive performance.

0 favorites 0 likes
#gradient-flow

GradCuit: Credit-Assigned Gradient Flow Enables Robust and Interpretable Test-Time Latent Reasoning

Hugging Face Daily Papers · 2026-08-03 Cached

This paper introduces GradCuit, a method for test-time latent reasoning that inserts optimizable latent states at a selected Transformer layer. It achieves 64.5% average accuracy across five backbones and three reasoning benchmarks, outperforming chain-of-thought prompting and showing improved robustness and interpretability.

0 favorites 0 likes
#gradient-flow

Gradient Flow Dynamics and Implicit Bias of Diagonal Linear Networks under Infinitesimal Initialization

arXiv cs.LG · 2026-07-15 Cached

This paper studies the gradient flow dynamics of diagonal linear networks under infinitesimal initialization, generalizing previous results to deep networks and a broader class. It shows that the implicit bias corresponds to a modified ℓ1 norm and identifies the Structural Invariant Manifold as a key geometric structure.

0 favorites 0 likes
#gradient-flow

@SebastienBubeck: https://x.com/SebastienBubeck/status/2075596982622835006

X AI KOLs Timeline · 2026-07-10 Cached

GPT-5.6 significantly outperforms published state-of-the-art on a fundamental mathematical problem about gradient flow length, achieving exponential improvements. This marks a major advance in AI's ability to reason about complex mathematical questions.

0 favorites 0 likes
#gradient-flow

A prism hierarchy of learning regimes in large linear autoencoders

arXiv cs.LG · 2026-06-05 Cached

This paper systematically identifies all qualitatively different extreme learning regimes for large weight-tied linear autoencoders, deriving explicit loss evolutions for five regimes associated with the faces of a triangular prism.

0 favorites 0 likes
#gradient-flow

Reinforcement Learning via Value Gradient Flow

Hugging Face Daily Papers · 2026-04-15 Cached

Value Gradient Flow (VGF) presents a scalable approach to behavior-regularized reinforcement learning by formulating it as an optimal transport problem solved through discrete gradient flow, achieving state-of-the-art results on offline RL and LLM RL benchmarks. The method eliminates explicit policy parameterization while enabling adaptive test-time scaling by controlling transport budget.

0 favorites 0 likes
← Back to home

Submit Feedback