Continual-ARC benchmark: leading agents fail to learn from experience
hey_abusiddik · x · 2026-11-12
A new benchmark called Continual-ARC from Unify.ai delivers a sobering result: OpenClaw, Hermes Agent and Prime Agent all failed to memorize anything of value or generalize to new tasks. Fine-tuning a small 9B model's weights, by contrast, achieved both memorization and generalization. The practical upshot: most people re-explain the same work context to AI every day, and an assistant that actually learned from your work would save hours weekly.
More from coding & agent
- United Airlines to Ban Personal AI Agents From Booking Flights for Customers — Polymarket · 2026-11-12
- Agentic Search explained: a 4-search loop costs about 5x one search — and queries leak to more companies — adnan_hashmi · 2026-11-12
- AWS says thousands of autonomous AI agents wrote every line of AWS Emulator — SumitGup · 2026-11-12
- freeCodeCamp launches free 2-hour course on agentic coding with Codex — adnan_hashmi · 2026-11-12
- A criminally underrated resource comparing tool-calling reliability across models and providers — ricklamers · 2026-11-12
- Organizing 55k data points is easy: enterprise data prep for agents is the hard part — BenBajarin · 2026-11-12