LLaDA2.2 adds delete and insert to diffusion LLMs for 128K agent workflows
机器之心 · wechat · 2026-07-28
Ant Group’s LLaDA2.2 pushes diffusion language models from static generation into agentic workflows by adding delete and insert to the edit space, not just keep and substitute.
The report describes three big changes: Levenshtein-style editing with four atomic operations, an RL method driven by environment feedback for tool-use and multi-turn trajectories, and native 128K context plus a fixed-capacity block routing design for MoE inference. On benchmarks, LLaDA2.2-flash is roughly on par with comparable autoregressive models on agent tasks, beats them on some tool-use benchmarks, and reaches 1.64x the average BF16 throughput.
Related event: Ant's LLaDA2.2 Brings Diffusion LLMs to Long-Context Agents(2 posts)→
More from coding & agent
- Vercel engineer shuding reimagines UI as complex-valued functions: viewport 720+42i px — shuding · 2026-10-27
- shuding: integrate your UI over real traffic to get the exact average page users see — shuding · 2026-10-27
- shuding takes the Fourier transform of a responsive dashboard as a frequency signal — shuding · 2026-10-27
- shuding mathematically proves UIs never break across infinite prop combos via interval arithmetic — shuding · 2026-10-27
- Filter-then-locate beats whole-file prompts: testing 4 ways to update READMEs with LLMs — KhuyenTran16 · 2026-10-27
- Qiaomu Clipper launches site: open-source Chrome clipper that saves reading notes to Obsidian — vista8 · 2026-10-27