Beyond Reasoning Gains: Mitigating General-Capability Forgetting in Large Reasoning Models

19 June 2026 at 06:00 CEST