We spent 2 hours working in the future
Thomas Kwa describes a tabletop exercise where METR researchers simulated having access to ~200-hour time horizon AIs.
Community feed
evald.ai
A focused stream of recent stories from the sources curated for this community. Latest: We spent 2 hours working in the future, Arena Leaderboard: The Unbreakable Ranking System That’s Revolutionizing AI Model Evaluation - CryptoRank, and Measuring progress toward AGI: A cognitive framework. Page 5.
Arena Leaderboard: The Unbreakable Ranking System That’s Revolutionizing AI Model Evaluation - CryptoRank
Arena Leaderboard: The Unbreakable Ranking System That’s Revolutionizing AI Model Evaluation CryptoRank
Measuring progress toward AGI: A cognitive framework
Google DeepMind proposes a cognitive framework to evaluate AGI and launches a Kaggle hackathon to build capability benchmarks
AI benchmark numbers are meaningless — here’s what to look for instead - MakeUseOf
AI benchmark numbers are meaningless — here’s what to look for instead MakeUseOf
Global Standards, Local Ground Truths: Piloting Multilingual, Multimodal AI Safety Understanding in APAC - MLCommons
MLCommons is developing the AILuminate Culturally-Specific Multimodal Benchmark to close the AI performance and representation gap across APAC cultures, languages, and real-world use cases.
YOLO for the MLPerf Inference v6.0 Edge Suite - MLCommons
MLPerf Inference v6.0 upgrades its edge object detection benchmark from RetinaNet to YOLOv11, bringing modern real-time detection to standardized AI hardware evaluation
Qwen3.5-9B tops every AI benchmark right now, but that's not how you should pick a model - XDA
Qwen3.5-9B tops every AI benchmark right now, but that's not how you should pick a model XDA
Review of the Anthropic Sabotage Risk Report: Claude Opus 4.6
External review from METR of Anthropic's Sabotage Risk Report for Claude Opus 4.6
MiniMax M2.5 Sparks AI Benchmark Fraud Debate - AI CERTs
MiniMax M2.5 Sparks AI Benchmark Fraud Debate AI CERTs
Bringing Text-to-Video to MLPerf Inference v6.0 - MLCommons
MLCommons introduces the new Text-to-Video benchmark in MLPerf Inference v6.0, based on the Wan2.2-T2V-A14B-Diffusers model and validated using the VBench framework. Learn about the key architectural decisions, including the adoption of the SingleStream...
How Anthropic’s Claude Opus 4.6 Broke Its Own AI Benchmark - WinBuzzer
How Anthropic’s Claude Opus 4.6 Broke Its Own AI Benchmark WinBuzzer
What is Model Evaluation? - IBM
What is Model Evaluation? IBM
More stories
More stories load automatically as you scroll.