DiffResearch is an open multi‑agent deep research framework that replaces the traditional autoregressive language model backbone with a diffusion language model (dLLM) across all stages of research, from planning to writing. In experiments on 100 Deep Research Bench tasks, the diffusion backbone achieves roughly six times faster end‑to‑end latency, largely due to significant speedups in planning and long‑form generation. The framework also introduces dFast, a decomposition mode that reads intermediate denoising states to generate subqueries, yielding a median 1.99× speedup without compromising report quality and achieving the best overall score on the RACE rubric. whyItMatters":"The study demonstrates that diffusion language models can dramatically reduce latency in deep research systems, enabling faster interactive use while maintaining or improving output quality."
arXiv:2606. 19475v1 Announce Type: new Abstract: Large Language Models (LLMs) have revolutionized language modeling through autoregressive generation, enabling strong performance across a wide range of tasks.
Large language models increasingly rely on external tools to access up-to-date information, perform computation, and interact with the outside world. For autoregressive models, tool use naturally fits the generation process: the model emits a tool call, waits for the result, and then continues generating.
arXiv:2608. 10438v1 Announce Type: new Abstract: Large language models increasingly rely on external tools to access up-to-date information, perform computation, and interact with the outside world.
arXiv:2607. 12829v1 Announce Type: cross Abstract: Diffusion large language models (dLLMs) offer a theoretical advantage in parallel generation over standard autoregressive models.
arXiv:2508. 10875v3 Announce Type: replace-cross Abstract: Diffusion Language Models (DLMs) are rapidly emerging as a powerful and promising alternative to the dominant autoregressive (AR) paradigm.
The paper introduces PILL, a new infilling technique for diffusion language models that eliminates the need for a preset initial length and reduces inference overhead. PILL uses probing-based length-free decoding, cutting down on extra forward passes and speeding up generation. Experiments across five diffusion models and eight benchmarks show PILL outperforms the strongest baseline with higher pass rates and BLEU-2 scores while running 1.82× faster.
The paper introduces Stitching Noisy Diffusion Thoughts, a self‑consistency framework that transforms low‑cost diffusion‑sampled reasoning into a reusable pool of step‑level candidates. By sampling many diverse trajectories with a masked diffusion language model, scoring each intermediate step with a process reward model, and stitching the highest‑quality steps across trajectories, the method recomputes only the final answer, separating exploration from evaluation. Experiments on math reasoning benchmarks show that step‑level recombination improves accuracy by up to 23.8% on six tasks while reducing latency by up to 1.8× compared to traditional diffusion models and unified architectures.
arXiv:2606. 19005v1 Announce Type: cross Abstract: Diffusion models have become a promising alternative to autoregressive models.
arXiv:2610.14162v1 Announce Type: cross Abstract: Diffusion language models (DLMs) offer parallel sampling for faster generation than autoregressive models, an increasingly important advantage as inf...
LADDER is a new framework that combines diffusion language modeling with Graph Retrieval-Augmented Generation (GraphRAG) to enable efficient multi‑hop reasoning. It introduces an event‑driven self‑clocking retrieval mechanism that triggers graph queries only when new entities appear, and an incomplete‑query graph propagation module that aggregates multi‑hop evidence during parallel decoding. Experiments on three multi‑hop QA benchmarks show that LADDER improves exact match from 39.6% to 45.2% while reducing latency by 4.1×.
arXiv:2603. 01331v3 Announce Type: replace-cross Abstract: Discrete diffusion language models (dLLMs) generate text by iteratively denoising a masked sequence.
The paper introduces the Denoising Workload Surface (DWS), a two‑dimensional probability surface that captures the block‑autoregressive generation structure of diffusion large language models (dLLMs). By preserving both output block and within‑block denoising step information, DWS enables a lightweight, prompt‑only predictor to estimate per‑request inference cost accurately, even on a single CPU core. In real‑world serving experiments, DWS reduces cost‑prediction error by up to 2.5× and improves end‑to‑end latency for online chatbots by up to 1.92×.