paper-with-me

Papers

Path-Dependent Denoising: A Non-Conservative Field Perspective on Order Collapse in Diffusion Language Models

2026-05-10 · Jeonseong Kim arxiv

Diffusion language models (DLMs) offer a structural alternative to autoregressive generation: denoising can update tokens in arbitrary orders or in parallel rather than along a fixed left-to-right chain. In practice, fast DLM decoding remains strongly order-sensitive and often drifts toward autoregressive-like trajectories. We trace this tension to compatibility. At each reverse-time step, a DLM provides local denoising conditionals over the unresolved tokens. Arbitrary-order denoising becomes well defined when these local conditionals compose into order-invariant pseudo-joints. We formalize this view by defining order-induced pseudo-joints and a local denoising circulation: the log-ratio between the two pseudo-joints obtained by swapping a pair of unresolved positions. This circulation is zero under compatible conditionals, and global order gaps decompose into sums of local circulations along adjacent swaps. We further separate incompatibility-driven path dependence from conditional-dependence error in parallel updates and from order-specific estimation error. The resulting framework provides inference-only diagnostics for testing when DLM decoding is genuinely order-free.

📄 PDF Abstract BibTeX arXiv:2605.09303

Code (0)

등록된 구현이 없습니다.

Similar Papers 제목 키워드 기반

Drifting Fields are not Conservative

2026-04-07 · Leonard T. Franz, Sebastian Hoffmann, Tim Weiland, Bernhard Schölkopf 외 arxiv

Drifting models have recently gained attention for generating high-quality samples in a single forward pass. During training, they learn a push-forward map by following a vector-valued field, the drift field. We ask whet…

Conservativeness of untied auto-encoders

2015-06-25 · Daniel Jiwoong Im, Mohamed Ishmael Diwan Belghazi, Roland Memisevic

We discuss necessary and sufficient conditions for an auto-encoder to define a conservative vector field, in which case it is associated with an energy function akin to the unnormalized log-probability of the data. We sh…

DecoderDenoising

On gauge freedom, conservativity and intrinsic dimensionality estimation in diffusion models

2024-02-06 · Christian Horvat, Jean-Pascal Pfister

Diffusion models are generative models that have recently demonstrated impressive performances in terms of sampling quality and density estimation in high dimensions. They rely on a forward continuous diffusion process a…

DenoisingDensity Estimation

A One-Size-Fits-All Solution to Conservative Bandit Problems

2020-12-14 · Yihan Du, Siwei Wang, Longbo Huang

In this paper, we study a family of conservative bandit problems (CBPs) with sample-path reward constraints, i.e., the learner's reward performance must be at least as well as a given baseline at any time. We propose a O…

AllMulti-Armed Bandits

The structure of conservative gradient fields

2021-01-03 · Adrian Lewis, Tonghua Tian

The classical Clarke subdifferential alone is inadequate for understanding automatic differentiation in nonsmooth contexts. Instead, we can sometimes rely on enlarged generalized gradients called "conservative fields", d…