Merge pull request #34 from redhat-et/feat/shared-ranking-benchmark
Taylor Agarwal
Unverified signature
295066d
feat: Move benchmark selection to shared Rankings controls
15 days ago
fix: deduplicate Results Explorer filter choices
CP Legendre
18fc065
feat: unify benchmark selection across ranking charts
CP Legendre
8e41b63
test: use eligible benchmarks in harness ranking fixture
CP Legendre
aa954a8
24 days ago
feat: benchmark whitelist for rankings
Daniel Blei
Unverified signature
707d8d5
28 days ago
feat: add SRE category with ITBench AA results
Daniel Blei
Unverified signature
8ab979c
- Add SRE tab to the leaderboard with benchmark section rendering
- Catalog ITBench AA as an SRE benchmark (offline, incident-response)
- Add ITBench AA results for GLM 5.3, Qwen3.8 Flash Next, and Sonnet 5
- Add ITBench AA results for Opus 5.5 and Haiku 4.5 (Claude Code)
- Regenerate results.csv to include new runs
- Ignore local egg-info build artifacts
- Bump version to 0.2.0
Reduce ranking row height
CP Legendre
38eb34d
Strengthen pairwise ranking regression assertions
CP Legendre
e21dd2d
Merge main and address ranking review feedback
CP Legendre
c797253
Scale dense charts with result count
CP Legendre
f1ae266
Rename metric-aware ranking order controls
CP Legendre
48c3ff8
Refine paired comparisons and ranking order
CP Legendre
0df8ce4
Add configuration metadata and comparison foundations
CP Legendre
62e4b94
Add efficiency, coding, generalist, and benchmark matrix views (#25)
Cédric P. Legendre+1: rounakbende
Unverified signature
7ddc8cb
* Add analytics views and benchmark matrices
* Add Bradley-Terry rankings tab with filters for benchmarks, models, harnesses, and rank-by metric
* Address CR: fix eqvdata population, empty filter handling, and per-table error isolation
* Address leaderboard review feedback
* Round trade-off table values to one decimal
* Round trade-off table values to one decimal
* Round trade-off metrics to two decimals
* Round trade-off metrics to two decimals
---------
Co-authored-by: rounakbende10 <rounakbende@gmail.com>