y0news
← Feed
Back to feed
🧠 AI🟢 BullishImportance 7/10

KARL: Knowledge Agents via Reinforcement Learning

arXiv – CS AI|Jonathan D. Chang, Andrew Drozdov, Shubham Toshniwal, Owen Oertell, Alexander Trott, Jacob Portes, Abhay Gupta, Pallavi Koppol, Ashutosh Baheti, Sean Kulinski, Ivan Zhou, Irene Dea, Krista Opsahl-Ong, Simon Favreau-Lessard, Sean Owen, Jose Javier Gonzalez Ortiz, Arnav Singhvi, Xabi Andrade, Cindy Wang, Kartik Sreenivasan, Sam Havens, Jialu Liu, Peyton DeNiro, Wen Sun, Michael Bendersky, Jonathan Frankle|
🤖AI Summary

Researchers present KARL, a reinforcement learning system for training enterprise search agents that outperforms GPT 5.2 and Claude 4.6 on diverse search tasks. The system introduces KARLBench evaluation suite and demonstrates superior cost-quality trade-offs through multi-task training and synthetic data generation.

Key Takeaways
  • KARL achieves state-of-the-art performance on enterprise search tasks, surpassing GPT 5.2 and Claude 4.6.
  • KARLBench introduces six distinct search evaluation regimes including constraint-driven search and cross-document synthesis.
  • Multi-task training across heterogeneous search behaviors provides better generalization than single-task optimization.
  • The system uses iterative bootstrapping with synthetic data generation for training high-quality knowledge agents.
  • KARL demonstrates Pareto-optimal performance across cost-quality and latency-quality trade-offs.
Mentioned in AI
Models
GPT-5OpenAI
ClaudeAnthropic
Read Original →via arXiv – CS AI
Act on this with AI
Stay ahead of the market.
Connect your wallet to an AI agent. It reads balances, proposes swaps and bridges across 15 chains — you keep full control of your keys.
Connect Wallet to AI →How it works
Related Articles