Unverified50% confidenceOpinionExact time
Bijective tasks are not uniformly harder than non-bijective OOD tasks and are even farther from the non-bijective OOD tasks used in the final evaluation, yet co-training remains effective, which contradicts the intuition that "the closer to the evaluation task, the more helpful."
1
Sources
50%
Confidence
Long-term
Relevance
11/3/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedTask-specific verifiers typically perform well within their training distribution but quickly fail when switched to a different task category, showing weak generalization ability.78% similarUnverifiedSimply stacking the number of RL tasks does not lead to better training outcomes, and unsolvable tasks instead drag down overall training quality.77% similarUnverifiedAt present, the collaboration between the two agents still relies on staged training rather than end-to-end joint optimization.74% similarUnverifiedSpecialized training for long-horizon tasks typically requires constructing supervised data containing complete task trajectories, rather than simple single-turn question-answer pairs.74% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1130331API
curl https://kongchang.com/api/v1/knowledge/claims/1130331MCP
get_claim(id=1130331)