TENCENT OPEN-SOURCES ANGELSPEC: A UNIFIED TRAINING FRAMEWORK FOR MTP AND BLOCK-PARALLEL SPECULATIVE DECODING ON HY3 MODELS

Tencent has released AngelSpec, an open-source torch-native framework for training speculative-decoding draft models across six architectures. It introduces DFly, a block-diffusion drafter with hybrid target conditioning and a hidden-correction autoregressive head, and integrates D-cut for runtime-adaptive verification budgeting.

MarkTechPost reports that tencent has released AngelSpec, an open-source torch-native framework for training speculative-decoding draft models across six architectures.

The report adds: It introduces DFly, a block-diffusion drafter with hybrid target conditioning and a hidden-correction autoregressive head, and integrates D-cut for runtime-adaptive verification budgeting.

On HY3-295B-A21B with TP=8, DFly-8 delivers a 1.98–2.40× speedup over autoregressive decoding across concurrency 4 to 64.

This report is covered across the current DAMMNEWS feeds. The most useful way to read a developing report is to separate the confirmed detail from early claims, then follow the next named source or decision.

This is a developing report. DAMMNEWS will surface related coverage as closely matched reporting enters the live cache.

DAMMNEWS RSS BRIEFING — generated locally from the available MarkTechPost headline and RSS excerpt. It is not a summary of the full source article.