claude plugin eval scores realistic prompts with 6 grader types; 4 are free, llm and baseline bill a judge model. Every case ...

Meta FAIR's AI Research Preference Models rank unexecuted ML candidates, raising AIRS-Bench from 0.684 to 0.729 without ...

Google releases TimesFM-3, a 330M parameter zero-shot foundation model for multivariate time series forecasting in one ...

Benchmarking the lowest-latency inference APIs for voice agents: measured TTFT, time to first audio, and full-pipeline ...

Google introduces EnvHarness, a programmable layer that reshapes static LLM agent environments without modifying their code.

Keenable AI open-sources NEEDLE, a live benchmark that rebuilds search queries hourly and scores 15 APIs for agents.

Princeton, Ant Group and Stanford built AQuA, two self-improving quant research agents whose sealed sandbox makes data leakage unwritable ...

Long-horizon agents have turned LLM serving into an input-heavy workload. Repeated prefills and million-token contexts leave KV caches that strain HBM, SSD capacity, and bandwidth. DeepSeek AI built ...

OpenBMB has released MiniCPM5-2B, the second checkpoint in the MiniCPM5 series and the follow-up to MiniCPM5-1B. It is a dense causal language model with 2,516,756,480 parameters, of which ...

ByteDance Seed, SUTD, Georgia Tech, M-A-P, and TokenWave.AI introduce HarnessDev, a benchmark that scores the runnable harness a model builds rather than the answer it returns. Starting from a seed ...

Multi-agent workflows have changed the shape of local inference. A lead agent decomposes a task and spawns subagents. What looked like one user request becomes dozens of independent model calls.

AI weather models have spent three years closing the gap with physics-based forecasting, but two problems stayed open: resolution too coarse for local terrain, and initialization tied to numerical ...