Self-Generated Error Training for Token Editing in Diffusion Language Models

Researchers propose self-generated token-to-token (T2T) editing for diffusion language models, addressing a training-inference mismatch where the editor sees the model's own fluent draft errors instead of random corruptions. The method improves accuracy and reduces edit intensity on benchmarks, mitigating failure modes like final-digit transcription errors.

arXiv:2606.17175v1 Announce Type: new Abstract: Token-to-token T2T editing lets LLaDA2.1 revise committed tokens during block-diffusion decoding. The released recipe trains this editor on random vocabulary corruptions, but at inference the editor sees the model's own fluent, high-confidence draft errors instead. We study this training-inference mismatch and propose self-generated T2T, which performs a no-gradient draft pass, fills masked positions with predicted tokens, and supervises recovery in a second pass under these self-generated corruptions. We implement the update as a short LoRA continued-pretraining pass on LLaDA2.1-mini and evaluate on several benchmarks under the official Q-Mode T2T procedure with unchanged inference parameters. The method generally improves accuracy while reducing T2T edit intensity, mitigating failure modes such as final-digit transcription errors after otherwise correct reasoning and excessive self-correction before short factual answers.