Работы о diffusion LM за 2026 год

Не полный каталог: здесь показаны профильные публикации, сохранённые в публичном корпусе. Число работ не измеряет качество или важность метода.

UTC: .

Профильных работ по текущим фильтрам: 9.

По месяцам и темам
МесяцРабот
Январь0
Февраль0
Март0
Апрель0
Май0
Июнь9
Июль0
Август0
Сентябрь0

Темы могут пересекаться: одна работа учитывается в каждой своей теме, но только один раз в общем числе. Связь с темой не доказывает качество или воспроизводимость.

Свежие работы · Методика отбора

2026-06-30T05:56:18+00:00 · Маскированная / дискретная диффузия, Рассуждения, Дообучение · Источник

Оригинальное название: TAG-DLM: Diffusion Language Models for Text-Attributed Graph Learning

Оригинальная аннотация: Text-attributed graphs (TAGs), where each node carries a natural language description, require models to jointly reason over text and graph topology. Existing approaches often handle the two modalities separately: graph neural networks operate on shallow text features, while hybrids of LLMs and graphs use the language model mainly as a text encoder and delegate structure learning to a separate graph module. We propose method that unifies textual reasoning and graph message passing within a masked diffusion language model, a language model with bidirectional attention and generative decoding. F

Полезно для: Не указано · Ограничение: Не указано

2026-06-28T05:53:45+00:00 · Маскированная / дискретная диффузия, Ускорение инференса, Дообучение · Источник

Оригинальное название: Multi-Block Diffusion Language Models

Оригинальная аннотация: Block Diffusion Language Models (BD-LMs) improve diffusion-based text generation with KV caching and flexible-length generation. A natural next step is to extend them from Single-Block Diffusion (SingleBD) to Multi-Block Diffusion (MultiBD), where a running-set of consecutive blocks is decoded concurrently for inter-block parallelism. However, existing BD-LMs are mostly trained under teacher forcing, where the model observes only one noisy block conditioned on a clean prefix. While the recent diffusion forcing strategy introduces visibility among multiple noisy blocks, its training states stil

Полезно для: Не указано · Ограничение: Не указано

2026-06-22T16:32:25+00:00 · Маскированная / дискретная диффузия, Рассуждения, Дообучение · Источник

Оригинальное название: Scheduling Thoughts: Learning the Order of Thought in Diffusion Language Models

Оригинальная аннотация: Masked diffusion language models decode by iteratively unmasking tokens, where the unmasking order defines an "order of thought" that strongly influences generation quality yet is typically chosen heuristically. We derive a tractable upper bound on the sequential decoding mismatch, measured by the Kullback-Leibler divergence and expressed in terms of the model's pathwise log-likelihood, with tightness under sufficient model expressivity. This bound induces a dense self-aware reward over ordered trajectories, casting order selection as a principled policy optimization problem wit

Полезно для: Не указано · Ограничение: Не указано

2026-06-10T16:14:23+00:00 · Рассуждения, Дообучение · Источник

Оригинальное название: Beyond Fully Random Masking: Attention-Guided Denoising and Optimization for Diffusion Language Models

Оригинальная аннотация: Diffusion large language models (dLLMs) offer an efficient alternative to autoregressive models through parallel decoding, yet existing post-training methods largely rely on random masking strategies that overlook intrinsic token dependencies. In this work, we present an empirical analysis of attention in dLLMs and show that tokens attending more strongly to unmasked context exhibit greater generation stability and play a critical role in reasoning. Motivated by these findings, we propose AGDO, an attention-guided denoising and optimization framework that aligns both training and optimization

Полезно для: Не указано · Ограничение: Не указано

2026-06-07T07:59:55+00:00 · Рассуждения, Дообучение · Источник

Оригинальное название: Back on Track: Aligning Rewards and States for Reasoning in Diffusion Large Language Models

Оригинальная аннотация: Reinforcement learning (RL) holds immense promise for enhancing the reasoning capabilities of diffusion large language models (dLLMs). However, progress is fundamentally constrained by a dual misalignment between authentic generation trajectory and the gradient update process: (i) Process-reward misalignment. Sparse, terminal rewards are indiscriminately assigned to all intermediate steps of the generation process, failing to provide discriminative credit assignment. (ii) State-trajectory misalignment. Policy updates are often diverted toward artificial, out-of-trajectory states, squandering g

Полезно для: Не указано · Ограничение: Не указано

2026-06-04T20:58:08+00:00 · Дообучение · Источник

Оригинальное название: Data-Efficient Autoregressive-to-Diffusion Language Models via On-Policy Distillation

Оригинальная аннотация: We study the transformation of autoregressive models (ARLMs) into diffusion language models (DLMs). Rather than pretraining from scratch, prior work replaces the causal attention in ARLMs with bidirectional attention and then trains the resulting model using a DLM objective. However, these approaches incur two distribution shifts. First, transitioning from a next-token prediction objective to a DLM objective can discard knowledge acquired by the ARLM during training. Second, standard DLMs suffer from a train-inference mismatch, as the training loss is defined on randomly masked sequences rathe

Полезно для: Не указано · Ограничение: Не указано

2026-06-04T08:00:51+00:00 · Маскированная / дискретная диффузия, Дообучение · Источник

Оригинальное название: FAIR-Calib: Frontier-Aware Instability-Reweighted Calibration for Post-Training Quantization of Diffusion Large Language Models

Оригинальная аннотация: Diffusion Large Language Models (dLLMs) refine tokens iteratively but commit them irreversibly, leading to a "stability lag" where early decisions remain fragile even after being written. We reveal that Post-Training Quantization (PTQ) error easily flips these borderline decisions at the write frontier, which are then permanently locked in and amplified. To address this, we propose Frontier-Aware Instability-Reweighted Calibration (FAIR-Calib), a two-stage PTQ framework for dLLMs. Stage I probes a full-precision teacher to estimate a position prior that combines frontier hits and mas

Полезно для: Не указано · Ограничение: Не указано

2026-06-03T14:34:35+00:00 · Дообучение · Источник

Оригинальное название: STaR-Quant: State-Time Consistent Post-Training Quantization for Diffusion Large Language Models

Оригинальная аннотация: Diffusion large language models (DLLMs) have recently emerged as a promising alternative to autoregressive LLMs by generating text through iterative masked denoising with bidirectional context. However, their large model sizes and iterative denoising process introduce substantial memory and computational overhead, motivating post-training quantization for efficient deployment. In this paper, we identify two key challenges for low-bit DLLM quantization: state-dependent activation disparity and temporal error accumulation. Masked and unmasked tokens exhibit different activation distributions wit

Полезно для: Не указано · Ограничение: Не указано

2026-06-03T14:34:35+00:00 · Дообучение · Источник

Оригинальное название: STaR-Quant: State-Time Consistent Post-Training Quantization for Diffusion Large Language Models

Оригинальная аннотация: Diffusion large language models (DLLMs) have recently emerged as a promising alternative to autoregressive LLMs by generating text through iterative masked denoising with bidirectional context. However, their large model sizes and iterative denoising process introduce substantial memory and computational overhead, motivating post-training quantization for efficient deployment. In this paper, we identify two key challenges for low-bit DLLM quantization: state-dependent activation disparity and temporal error accumulation. Masked and unmasked tokens exhibit different activation distributions wit

Полезно для: Не указано · Ограничение: Не указано

Даты публичного снимка
Сбор, зафиксированный в снимке
Самая новая публикация по данным снимка
Построение снимка

Это сохранённые сведения, а не время последней попытки сборщика или гарантия полноты корпуса.