SeLMRoute routes LLM queries via probabilistic semantic evidence, hitting 72.08% accuracy
Indigma · hf · 2026-10-01
SeLMRoute is a new LLM routing framework that decouples candidate-independent semantic evidence extraction from performance learning and deployment objectives.
- A decision model answers interpretable questions about each query (reasoning needs, external knowledge), keeping each judgment as a probability distribution.
- A lightweight supervised router uses this probabilistic semantic state to estimate candidate performance, then applies performance- or cost-aware objectives.
- On LLMRouterBench (15 datasets, 20 models, 11,481 queries), it reaches 72.08% accuracy vs 69.23% for the strongest fixed candidate; in a 13-model performance-cost setting it improves all five grouped splits with mean PerfGain of 2.66%.
Code: github.com/Indigma-Innovations/SeLMRoute
More from Research
- Custom CUDA SGEMM kernel matches cuBLAS, wins on 16384x16384+ matrices — goyal__pramod · 2026-10-02
- Mathematician Kevin Buzzard: math community is 'grieving' through five stages over AI — littmath · 2026-10-02
- AI runs DeepMind's consciousness calculator on itself: scores swing 0.10–0.72 by theory — izzycognita · 2026-10-02
- MINTEval, accepted at NeurIPS, shows LLM agents fail at tracking evolving contexts — EliasEskin · 2026-10-02
- MazeBench goes free online: agents prefer top-down views and stall in 3D puzzles — patience_cave · 2026-10-02
- Aleph Alpha Details MoE Pre-training Scaling: 30B-A3B on 16-512 B200s at 35% MFU — CatAstro_Piyush · 2026-10-02