arXiv AI July 1, 2026 By Zhongyao Wang

Why Do Few-Step Text Latents Fail When Image Latents Work? Non-Commitment at Sharp Categorical Readouts

arXiv:2606. 30705v1 Announce Type: cross Abstract: Deterministic few-step generation succeeds on continuous image latents but collapses to incoherent text on continuous text latents, and we show the cause is geometric rather than a training or scaling deficiency: a smooth, regularity-limited deterministic map cannot resolve a discrete branch choice before a sharp categorical readout, so few-step failure is governed by decoder sharpness, not transport accuracy.

Machine-generated by The Flow from the publisher's headline and feed description — not written or checked by a human. The full article lives at arXiv AI.

arXiv Machine Learning

Aug 12

Intrinsic Structure: Spectral Identifiability for Mechanistic Interpretability

arXiv:2608. 10172v1 Announce Type: new Abstract: Mechanistic interpretability explains models by identifying circuits inside them, but has no way to tell whether a circuit is a property of the model or an artifact of the method that found it.

By Ashim Dhor, Pin-Yu Chen

llms safety

arXiv AI

Jul 3

Expander Sparse Autoencoders: Parameter-Efficient Dictionaries for Mechanistic Interpretability

arXiv:2607. 01799v1 Announce Type: cross Abstract: Sparse autoencoders (SAEs) decompose internal activations of neural networks into sparse linear combinations of learned features by fitting an overcomplete dictionary $\mathbf{W}\in\mathbb{R}^{m\times n}$ with $m<n$, and inferring a sparse code $\mathbf{x}\in\mathbb{R}^n$ from $\mathbf{h}\approx\mathbf{W}\mathbf{x}$.

By Rodrigo Mendoza-Smith

llms safety

arXiv AI

Jul 29

Reading Without a Reader: Large Language Models Collapse Reading and Writing into a Single Entangled Code

arXiv:2607. 24797v1 Announce Type: cross Abstract: In the literate human brain, reading and writing are two doubly-dissociable systems: a ventral decoding route (impaired in pure alexia) and a fronto-parietal encoding route (impaired in pure agraphia), sharing a partial orthographic core.

By Diego Salda\~na Ulloa

llms

arXiv Machine Learning

Jul 7

Asymptotic-Preserving A Posteriori Analysis of Diffusion and Flow-Matching Samplers

arXiv:2607. 04113v1 Announce Type: new Abstract: Diffusion and flow-matching samplers integrate a learned probability-flow ODE from a large noise scale down to a small terminal floor $\sigma_{\min}$, at which the score is stiff and the flow develops a boundary layer.

By Shiheng Zhang

diffusion

arXiv Machine Learning

Aug 27

Drift Variation Autoencoder: Unifying Generation and Representation Learning through Conditional Posterior Flow Matching

arXiv:2608. 25138v1 Announce Type: new Abstract: Stochastic masking, cropping, or modality removal makes deterministic reconstruction an incomplete target: one observation can admit many clean completions.

By Jiarui Cao

diffusion multimodal benchmarks

arXiv Machine Learning

Jun 2

Capability and Robustness Cannot Both Be Free: An Information-Theoretic Bound for Vision-Language-Action Models

arXiv:2605. 25889v4 Announce Type: replace-cross Abstract: Vision-Language-Action (VLA) models reach high success rates on clean inputs but collapse under small adversarial perturbations: a $16/255$ PGD attack drops OpenVLA-7B's LIBERO success from $95\%$ to under $5\%$.

By Jianwei Tai

multimodal safety

One email a morning, machine-written

Related stories

Intrinsic Structure: Spectral Identifiability for Mechanistic Interpretability

Expander Sparse Autoencoders: Parameter-Efficient Dictionaries for Mechanistic Interpretability

Reading Without a Reader: Large Language Models Collapse Reading and Writing into a Single Entangled Code

Asymptotic-Preserving A Posteriori Analysis of Diffusion and Flow-Matching Samplers

Drift Variation Autoencoder: Unifying Generation and Representation Learning through Conditional Posterior Flow Matching

Capability and Robustness Cannot Both Be Free: An Information-Theoretic Bound for Vision-Language-Action Models