Berkeley AI Research proposes GRASP, a gradient-based world model planner that makes long-horizon planning more practical. Three core improvements: lifting trajectories into virtual states so optimization runs in parallel over time, injecting randomness directly into state iterations to support exploration, and reshaping gradients to avoid traversing fragile state-input gradients of high-dimensional vision models.
Google Research introduced TurboQuant, a compression algorithm (to be published at ICLR 2026) that combines Quantized Johnson-Lindenstrauss (QJL) and PolarQuant (to be published at AISTATS 2026) to achieve zero-memory-overhead vector compression, compressing models and KV cache without accuracy loss.
Mistral AI released Mistral Small 4, the first small model to unify Magistral reasoning, Pixtral multimodal and Devstral agentic coding capabilities into a single model, open sourced under the Apache 2.0 license.
Why it matters: The official release provides full architecture parameters, benchmark comparisons and deployment requirements, letting readers assess whether it can replace a multi-model stack.
Google Research, with Cornell and Harvard, published a PNAS paper in which 12 high-temperature superconductivity experts wrote 67 questions and blind-reviewed answers from 6 LLMs.
Mistral has released its new Mistral 3 model series, including the MoE-architecture Mistral Large 3 (41B active, 675B total parameters) and Ministral 3 models in 3B, 8B, and 14B sizes, all open sourced under Apache 2.0.
Why it matters: The official announcement fully lists model specs, open source licenses, deployment options, and available platforms, so readers can assess feasibility of running them in their own environments.
Berkeley AI Research proposes an off-policy RL value-learning algorithm based on divide and conquer that no longer uses temporal difference (TD) learning, whose Bellman recursion error accumulation makes it hard to scale to long-horizon tasks.
Berkeley AI Research proposes a quantitative theory of the word2vec learning process: under several mild approximations, its learning problem reduces to unweighted least-squares matrix factorization, the gradient flow dynamics have a closed-form solution, and the final embeddings are equivalent to PCA on the target matrix M*.
Mistral AI released its first reasoning model, Magistral, in two versions: the open-source Magistral Small (24B, Apache 2.0) and the enterprise Magistral Medium.
Why it matters: With both open-source and commercial versions, official AIME2024 scores, open licensing and deployment channels, readers can use this to evaluate model selection.