Artificial Analysis launched VIALS-AA, a leaderboard built on Handshake AI's VIALS benchmark, evaluating frontier models on 161 life-science visual reasoning tasks. Gemini 4 Carbon leads narrowly at 64% pass^3, followed by GPT-6 Astra at 63% and Claude Opus 5.5 at 61%, a 3-point spread. Refusals on several tasks show scientific artifact interpretation remains a clear weakness for frontier models.
Pass@k
Metric measuring the probability that at least one of k generated samples is correct.
Curated Intelligence
Researchers from Shanghai Institute of AI and others proposed PSN-RLVR, which applies consistent small random perturbations to model parameters throughout the reasoning trajectory. This method addresses the issue where RL-trained models become accurate but lose diversity in solution paths, significantly improving Pass@K success rates and out-of-distribution generalization. The paper has been accepted by NeurIPS 2026.
To address the failure of standard RLVR in tasks with zero success rates, this paper introduces RLTL;DR, a framework where agents generate concise 'TL;DR' insights after failures to guide subsequent rollouts. By backpropagating on these in-context insights, the model internalizes a direct task-to-insight mapping. Experiments show this method boosts Qwen 3.5 9B's Pass@1 from near 0% to 12-13% on extremely difficult datasets, achieving significant performance recovery with minimal training data.
Moonshot AI released Kimi K2-Instruct, demonstrating superior performance in coding, agentic tool use, and complex reasoning benchmarks. Data indicates it matches or outperforms leading open-source and proprietary models like DeepSeek-V3, Claude Sonnet 4, and GPT-4.1, with notable breakthroughs on SWE-bench Verified and LiveCodeBench.