🧪 Test?View on arXiv
FusionEval: Benchmarking Model Fusion for Large Language Models
Baicai Haochi, Author 2, Author 3, Author 4, Author 5
model fusionbenchmarkingexpert modelstool use
2610.30162
Builder Relevance
2h ago80%
Abstract
FusionEval introduces a benchmark for evaluating model fusion methods that integrate multi-expert capabilities while addressing reasoning and tool use.
Reality Card
Core Claim
FusionEval demonstrates that training-free fusion methods preserve reasoning performance but degrade tool use, while MOPD maintains tool use with a balanced capability profile at the cost of additional data preparation and training.
Method / Result
MOPD achieves a more balanced capability profile compared to training-free methods.
Limitations
The requirement for additional data preparation, inference, and training for MOPD may hinder reproducibility.
Paper to code
Verified implementation resources so builders can test the paper’s claims instead of stopping at the abstract.
No verified implementation link has been attached yet. AIBuzzHub will keep this panel separate from unverified search results.