Probing Minimalist Phase Structure in LLMs: What Universal Dependencies Cannot Represent
- Published
- Source
- arXiv
- Paper number
- 246
- Field
- LLMs / NLP
- arXiv ID
- 2605.26431
Key points
- Use wh-movement stimuli designed so that UD tree distance remains invariant across conditions to detect structure beyond UD.
- In 12 of 13 LLMs across four families, the authors find a distance slope proportional to Minimalist phase count, with βfin > βinf > 0.
- Discover 13 out of 13 sign asymmetries that exactly match within-phase cohesion predictions, a result impossible to capture with UD-based probing.
- Introduce canonical-layer reporting, which removes degrees of freedom from per-contrast peak reporting and applies a stricter standard.
- Confirm the causal relevance of probe results with activation patching in 12 of 13 models.
- The results suggest that distributional pretraining can induce formal syntactic representations that go beyond the limits of annotation-based probing.
Paper links
External research summaries. These are not HDATF publications or measured product results.