Characterize Then Distill: Mechanistic Reasoning in Large Output Spaces
About
Modern reasoning models offer surprisingly strong zero-shot performance on challenging multi-label tasks that require selecting a small set of relevant options from hundreds of thousands to millions of candidate labels. We investigate how they achieve this mechanistically. We characterize reasoning as a two-phase process: A broad "shortlisting" of candidates followed by fine-grained reasoning over the resulting set. We provide evidence across a range of datasets that these steps can be isolated and are complementary. Using this characterization, we develop a mechanistic distillation strategy that consistently outperforms standard distillation.
Debjyoti Saha Roy, Byron C. Wallace, Javed A. Aslam• 2026
Related benchmarks
| Task | Dataset | Result | Rank | |
|---|---|---|---|---|
| Multi-Label Classification | MIMIC-IV full | Foc0.66 | 13 | |
| Extreme Multi-Label Retrieval | LF-WikiSeeAlso-320K | Focus70 | 5 |
Showing 2 of 2 rows