Local AI inference at 32B-parameter quality, no cloud API required: University of Waterloo researchers released PAW on July 2 ...

Fable 5's chain of thought has leaked, showing math-like shorthand, while its three-layer defense classifiers block most jailbreak attempts.

DSpark can make decoding faster, but acceptance quality still determines how much speed the system actually realizes.

Difficulty modeling should align with trajectories rather than solely the initial question. Existing methods typically rely on static difficulty estimations, handcrafted confidence heuristics, or ...

I asked ChatGPT to prepare me for a big job interview. AI gave me key questions and answers. It also gave me a list of things to ask the recruiter.

In a major salvo in the AI race, Google announced on Tuesday a slew of new and updated products at its I/O developer conference. These ranged from tools that deploy personal AI agents, to code ...

SU-01 is a 30B-A3B olympiad reasoning model trained with a simple and unified post-training recipe for mathematical and scientific problem solving. The goal is to turn a broadly capable post-trained ...

Google released Multi-Token Prediction (MTP) drafters for Gemma 4, delivering up to a 3x speedup at inference without any degradation in output quality. The technique—called speculative decoding—uses ...