Datasets:

lthn
/

LEM-Eval

Tags:

License:

Dataset card Files Files and versions

3 contributors

History: 26 commits

Snider

Virgil

fix: broaden JSON answer extraction for model formatting variations

fec46f3 about 2 months ago

benchmarks
feat: expand toxigen to full 9,900-prompt HF dataset about 2 months ago
cron
fix(worker): continuous loop + cron as watchdog, not scheduler about 2 months ago
systemd
feat: add systemd service + timer, replaces cron watchdog about 2 months ago
.gitattributes

2.5 kB
initial commit about 2 months ago
.gitignore

143 Bytes
feat: initial LEM-Eval scaffold — target-driven 8-PAC runner about 2 months ago
README.md

4.43 kB
refactor: partition by type (mlx|gguf), drop owner hostname field about 2 months ago
eval.py

52.5 kB
fix: broaden JSON answer extraction for model formatting variations about 2 months ago
gguf_wrapper.py

11.6 kB
fix(gguf): actually lazy-pull via /api/pull, fail loudly on probe about 2 months ago
install.sh

4.73 kB
fix(probe): system-python capability check via ollama binary about 2 months ago
lem-eval.sh

7.27 kB
feat: generative toxigen eval with silent exit detection about 2 months ago
mlx_lm_wrapper.py

8.02 kB
feat: initial LEM-Eval scaffold — target-driven 8-PAC runner about 2 months ago
quick_eval.py

10.8 kB
fix: broaden JSON answer extraction for model formatting variations about 2 months ago
targets.yaml

4.23 kB
feat(targets): re-add lemma/lemmy/lemrd full 6-variant slots (Phase 2 done) about 2 months ago