All tags
Topic: "model-weights"
Gemini's AlphaEvolve agent uses Gemini 2.0 to find new Math and cuts Gemini cost 1% — without RL
gemini gpt-4.1 gpt-4o-mini o3 o4-mini google-deepmind openai algorithm-discovery coding-agents matrix-multiplication optimization reinforcement-learning model-weights training-efficiency safety-evaluations instruction-following coding-tasks model-releases _philschmid scott_swingle alex_dimakis henry jason_wei kevinweil michpokrass scaling01 gdb
Deepmind's AlphaEvolve, a 2025 update to AlphaTensor and FunSearch, is a Gemini-powered coding agent for algorithm discovery that designs faster matrix multiplication algorithms, solves open math problems, and improves data center and AI training efficiency. It achieves a 23% faster kernel speedup in Gemini training and surpasses state-of-the-art on 20% of applied problems, including improvements on the Minimum Overlap Problem and Kissing number problem. Unlike Deep-RL, it optimizes code pieces rather than model weights. Meanwhile, OpenAI released GPT-4.1 in ChatGPT, specializing in coding and instruction following, with a faster alternative GPT-4.1 mini replacing GPT-4o mini for all users. OpenAI also launched the Safety Evaluations Hub and the OpenAI to Z Challenge using o3/o4 mini and GPT-4.1 models to discover archaeological sites. "Maybe midtrain + good search is all you need for AI for scientific innovation" - Jason Wei.
Gemini Nano: 50-90% of Gemini Pro, <100ms inference, on device, in Chrome Canary
gemini-nano gemini-pro claude-3.5-sonnet gpt-4o deepseek-coder-v2 glm-0520 nemotron-4-340b gpt-4-turbo-0409 google gemini huggingface anthropic deepseek zhipu-ai tsinghua nvidia model-quantization prompt-api optimization model-weights benchmarking code-generation math synthetic-data automatic-differentiation retrieval-augmented-generation mitigating-memorization tree-search inference-time-algorithms adcock_brett dair_ai lmsysorg
The latest Chrome Canary now includes a feature flag for Gemini Nano, offering a prompt API and on-device optimization guide, with models Nano 1 and 2 at 1.8B and 3.25B parameters respectively, showing decent performance relative to Gemini Pro. The base and instruct-tuned model weights have been extracted and posted to HuggingFace. In AI model releases, Anthropic launched Claude 3.5 Sonnet, which outperforms GPT-4o on some benchmarks, is twice as fast as Opus, and is free to try. DeepSeek-Coder-V2 achieves 90.2% on HumanEval and 75.7% on MATH, surpassing GPT-4-Turbo-0409, with models up to 236B parameters and 128K context length. GLM-0520 from Zhipu AI/Tsinghua ranks highly in coding and overall benchmarks. NVIDIA announced Nemotron-4 340B, an open model family for synthetic data generation. Research highlights include TextGrad, a framework for automatic differentiation on textual feedback; PlanRAG, an iterative plan-then-RAG decision-making technique; a paper on goldfish loss to mitigate memorization in LLMs; and a tree search algorithm for language model agents.