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 training framework for speculative-decoding draft models. The release covers both autoregressive multi-token prediction (MTP) and the block-parallel DFlash family. Most speculative-decoding work searches for one drafter that scores well on an averaged benchmark mixture. Real serving traffic does not look like that mixture. AngelSpec treats workload heterogeneity as a first-class design constraint, and specializes structure, traini
Get smart on it
Tencent has released AngelSpec, an open-source training framework that improves speculative decoding, a technique where a lightweight model proposes multiple future tokens and a larger model verifies them together to speed up text generation. Rather than training a single universal drafting model, AngelSpec uses two complementary approaches: one model optimized for conversational text where many continuations are valid, and another designed for code and mathematics where token sequences are more predictable. The framework addresses a mismatch between how models are trained and how they are used at serving time by having drafters practice generating long chains during training with feedback from the actual target model's choices, which improves acceptance rates at deeper prediction positions.
Related news
Pokee AI Releases Pokee-Isaac 28B: A 10M-Token Context Agentic Model Built to Run Inside the Customer Boundary
Long-horizon agents accumulate context faster than they resolve tasks. Every tool output, observation, and intermediate reasoning step stays in the window, and the two capabilities that matter — holding that context and staying coherent across it — have so far been available almost exclusively from cloud endpoints. That excludes regulated industries, public-sector institutions, and on-device applications, where the data is not permitted to leave the boundary at all. Pokee AI released Pokee-Isaa
Mistral AI Releases Shieldstral 1.0 3B: An Open-Weights Policy-Adaptive Multimodal Safety Classifier Matching Models 7× Its Size
Mistral AI has released Shieldstral 1.0 3B, an open-weights, policy-adaptive multimodal safety classifier that treats content moderation as a single yes/no question rather than a fixed taxonomy of harm categories. Most guardrail models bake their category list into the weights, so re-targeting one to a new deployment context means retraining — and the same content can be acceptable on a cybersecurity research tool while being harmful on a mental-health platform. Shieldstral inverts that: operat
