On July 27 Ant Group launched LLaDA2.2, introducing Levenshtein editing and
environment-feedback reinforcement learning to enable diffusion language models
in real agent scenarios to perform online "act-and-correct" behavior.
2026-07-27
On July 27 Ant Group launched LLaDA2.2, introducing Levenshtein editing and
environment-feedback reinforcement learning to enable diffusion language models
in real agent scenarios to perform online "act-and-correct" behavior.
