| Name | Modified | Size | Downloads / Week |
|---|---|---|---|
| Parent folder | |||
| 2.13.6 source code.tar.gz | < 21 hours ago | 10.8 MB | |
| 2.13.6 source code.zip | < 21 hours ago | 13.1 MB | |
| README.md | < 21 hours ago | 1.2 kB | |
| Totals: 3 Items | 23.9 MB | 0 | |
2.13.6 (2026-05-17)
Fix
- fix: retrieval statistics calculation (#4671)
fix statistics calculation (3912be4)
Unknown
-
dataset: add HumanConceptsClustering (Shani et al. 2025) (#4609)
-
dataset: add HumanConceptsClustering from Shani et al. 2025
Adds HumanConceptsClustering task based on the cognitive psychology typicality datasets from Rosch (1973, 1975) and McCloskey & Glucksberg (1978), compiled in arXiv:2505.17117. Evaluates how well embeddings recover human conceptual organisation using k-means + V-measure across three sub-studies as separate splits (Rosch1973, Rosch1975, McCloskey1978).
Closes [#4497]
-
add statistics
-
lint
-
fix: add HumanConceptsClustering to KNOWN_ISSUES for short single-word items
-
feat: set AMI as main score for HumanConceptsClustering
AMI is the metric used in arXiv:2505.17117. The clustering evaluator computes both v_measure and ami; main_score selects ami.
Co-authored-by: Roman Solomatin <36135455+Samoed@users.noreply.github.com> (cc71dba)