RAGEN ranks 3rd hardest in FrontierCS-2 — built from code hand-written in Jan 2025
wzenus · x · 2026-10-30
wzenus discovered that RAGEN — turning multi-turn agent trajectories into training batches — ranks as the 3rd hardest task (FDI 92.9) among 48 tasks in FrontierCS-2, behind KV Compaction (107.0) and Structured LWE (101.4).
The core code was written entirely by hand in Jan 2025, before coding agents took off. The author muses at how quickly hand-written work became a frontier benchmark task.
More from Research
- Amazon's DreamFAR world model nearly doubles π₀ real-world performance for robot training — CyberRobooo · 2026-10-30
- Investor's Last Exam Benchmark: Top AI Models Ace Research but Fail at Investment Judgment — Huangfei Jiang · 2026-10-30
- SpecSlot Boosts Pixel-Space Image Generation: +2.84 FID on ImageNet-512 at Lower Cost — Jiongfan Yang · 2026-10-30
- Apple Trains UI Navigation Agents with Online RL Directly on Live iOS Apps — apple · 2026-10-30
- KAIST's L3G-World Generates Infinite Interactive Worlds in Real Time with Latent 3D Gaussian Memory — kaist-ai · 2026-10-30
- PhysInsert from Beihang and Meitu Achieves Physically Plausible Reference-Based Object Insertion — Yue Gong · 2026-10-30