Beyond Uncertainty: Multi-Solver Disagreement Rewards for Self-Evolving Reasoning Curricula

Sicherheitsmeldung mit Schweregrad noch nicht bewertet. Technische Details im Tab „Originaltext“; empfohlene Schritte in der Checkliste.

Originaltext der Quelle (nicht übersetzt) – die deutsche Zusammenfassung findest du im Tab „Kurzfassung".

arXiv:2608.30035v1 Announce Type: new Abstract: Self-evolving reasoning frameworks train a Challenger to generate questions exposing a Solver's weaknesses, creating adaptive curricula without human data. However, existing approaches use a single solver's sampling uncertainty as the Challenger's reward. This creates a fundamental bottleneck: as the solver grows confident on the Challenger's question distribution, all sampled answers converge identically, collapsing the reward to zero and starving the Challenger of learning signal. Critically, this single-model reward cannot distinguish genuinely easy questions from those that merely align with one solver's learned biases. We propose a multi-solver disagreement reward using a heterogeneous ensemble varying in model capacity and sampling temperature. A normalized Shannon entropy over the ensemble's per-question plurality answers explicitly rewards questions where solvers produce conflicting solutions---capturing difficulty as inter-model divergence rather than intra-model sampling variance. This richer gradient enables the Challenger to discover questions targeting true capability boundaries, producing a curriculum that forces downstream Solvers to develop robust reasoning strategies generalizing across problem types. Our approach is a drop-in reward function replacement requiring no framework modifications or additional data. Experiments with Qwen3-4B show that Solvers trained on disagreement-Challenger questions achieve +1.34 points average improvement on competition-math benchmarks (MATH-500, AMC, Olympiad), suggesting …

Vollständiger Originalartikel ↗

• Betroffenheit im eigenen Stack prüfen: Versionen/Komponenten abgleichen.
• Originalquelle / Hersteller-Advisory lesen: https://arxiv.org/abs/2608.30035
• Verfügbaren Patch oder Workaround einspielen und dokumentieren.