On this day: Jul 25
4 records from this date, 2023 to 2025, gathered year by year — the most recent year first.
2025
1 year ago · 1 recordGEPA publishedLakshya A Agrawal et al. (UC Berkeley, Stanford and others)
Argues that prompt evolution can compete with RL on sample efficiency
Prompt EngineeringPaper
2024
2 years ago · 2 recordsGoogle DeepMind announces AlphaProof and AlphaGeometry 2 results at IMO 2024Google DeepMind
Self-training grounded in a formal language reached olympiad level. DeepMind stated explicitly that it would not have been possible without Lean and Mathlib
Andrej Karpathy posts the term jagged intelligence on XAndrej Karpathy
Gave a name to the gap between benchmark scores and reliability in real use. Salesforce adopted it as a formal research agenda and built the SIMPLE benchmark, and Sundar Pichai riffed on it as AJI
People & DiscoursePress
2023
3 years ago · 1 recordWebArenaCMU (Zhou et al.)
In an evaluation on real, working websites, a GPT-4 agent scored 14.41% versus 78.24% for humans, nailing down the gap between demos and reality
Tools & AgentsPaper