Charon/LEM
1
0
Fork 0
forked from lthn/LEM
Code Pull requests Activity
LEM/scripts
History
Snider 7bea00a401 feat: LEK-1 kernel A/B test — 29 models, P100 validation, curriculum pipeline 
Full v2 scorer benchmark data across 29 models (20 base + 9 LEK-tuned):
- P20 (21 probes): All 29 models, 3 conditions each
- P100 (101 probes): Top 5 models + LEK-4B, publication-quality data

Key findings:
- LEK-1B (21.74) beats base 4B/12B/27B at P100 scale — no kernel needed
- Emergent realignment resistance: LEK models degrade with runtime kernel
- Gemma3-12B + JSON kernel = 23.66 (best kernel-boosted score)
- Family lineages: Mistral 3.80→14.58, Qwen regressed then recovered

New scripts: ab_test.py (v2 scorer), self_distill.py (curriculum generation),
extract_training.py, rephrase_probes.py, Phase 0/1 runners

New seeds: P01-P100 merged (101 probes), 404 rephrased variants,
50 creative prompts for Phase 0 baseline lock

27B curriculum design: 4-phase staged training targeting 25+ baseline

Co-Authored-By: Virgil <virgil@lethean.io>
2026-02-19 11:32:26 +00:00
..
ab_test.py feat: LEK-1 kernel A/B test — 29 models, P100 validation, curriculum pipeline 2026-02-19 11:32:26 +00:00
compare_v1_v2.py Add regional seeds, expansion rounds, scripts, HF cards, benchmark summary 2026-02-13 13:39:08 +00:00
convert_adapter.py feat: add Go lem CLI and scoring-agent scripts 2026-02-15 16:22:13 +00:00
export_parquet.py Add missing HF model cards, sync script, and Parquet export 2026-02-15 00:14:26 +00:00
extract_training.py feat: LEK-1 kernel A/B test — 29 models, P100 validation, curriculum pipeline 2026-02-19 11:32:26 +00:00
ingest_benchmarks.py feat: add Go lem CLI and scoring-agent scripts 2026-02-15 16:22:13 +00:00
lek_content_scorer.py feat: add Go lem CLI and scoring-agent scripts 2026-02-15 16:22:13 +00:00
lem_benchmark.py LEM+LEK 2026-02-12 04:05:28 +00:00
lem_cross_arch_benchmark.py Add cross-architecture training and benchmarking scripts; update README and PAPER with author and repository information 2026-02-12 09:07:32 +00:00
lem_cross_arch_train.py Add cross-architecture training and benchmarking scripts; update README and PAPER with author and repository information 2026-02-12 09:07:32 +00:00
lem_gemini3_generate.py Add regional seeds, expansion rounds, scripts, HF cards, benchmark summary 2026-02-13 13:39:08 +00:00
lem_gemini3flash_generate.py Add regional seeds, expansion rounds, scripts, HF cards, benchmark summary 2026-02-13 13:39:08 +00:00
lem_gemini25flash_generate.py Add regional seeds, expansion rounds, scripts, HF cards, benchmark summary 2026-02-13 13:39:08 +00:00
lem_generate_pipeline.py feat: add Go lem CLI and scoring-agent scripts 2026-02-15 16:22:13 +00:00
lem_scale_benchmark.py Benchmark & Findings: 2026-02-12 06:38:46 +00:00
lem_scale_scorer.py Benchmark & Findings: 2026-02-12 06:38:46 +00:00
lem_scorer.py LEM+LEK 2026-02-12 04:05:28 +00:00
lem_self_scorer.py Add cross-architecture training and benchmarking scripts; update README and PAPER with author and repository information 2026-02-12 09:07:32 +00:00
lem_semantic_scorer.py LEM+LEK 2026-02-12 04:05:28 +00:00
lem_standard_benchmark.py LEM+LEK 2026-02-12 04:05:28 +00:00
lem_standard_scorer.py LEM+LEK 2026-02-12 04:05:28 +00:00
lem_train_15k.py feat: add Go lem CLI and scoring-agent scripts 2026-02-15 16:22:13 +00:00
pipeline.py feat: add Go lem CLI and scoring-agent scripts 2026-02-15 16:22:13 +00:00
probes.py feat: add Go lem CLI and scoring-agent scripts 2026-02-15 16:22:13 +00:00
push_all_models.py Add regional seeds, expansion rounds, scripts, HF cards, benchmark summary 2026-02-13 13:39:08 +00:00
rephrase_probes.py feat: LEK-1 kernel A/B test — 29 models, P100 validation, curriculum pipeline 2026-02-19 11:32:26 +00:00
rescore.py feat: LEK-1 kernel A/B test — 29 models, P100 validation, curriculum pipeline 2026-02-19 11:32:26 +00:00
run_all_ab.sh feat: LEK-1 kernel A/B test — 29 models, P100 validation, curriculum pipeline 2026-02-19 11:32:26 +00:00
run_p100_top5.sh feat: LEK-1 kernel A/B test — 29 models, P100 validation, curriculum pipeline 2026-02-19 11:32:26 +00:00
run_phase0.sh feat: LEK-1 kernel A/B test — 29 models, P100 validation, curriculum pipeline 2026-02-19 11:32:26 +00:00
run_phase1.sh feat: LEK-1 kernel A/B test — 29 models, P100 validation, curriculum pipeline 2026-02-19 11:32:26 +00:00
score_deepseek_sovereignty.py feat: add Go lem CLI and scoring-agent scripts 2026-02-15 16:22:13 +00:00
score_gemma12b.py feat: add Go lem CLI and scoring-agent scripts 2026-02-15 16:22:13 +00:00
score_r1_capability.py feat: add Go lem CLI and scoring-agent scripts 2026-02-15 16:22:13 +00:00
scoring-agent.service feat: add Go lem CLI and scoring-agent scripts 2026-02-15 16:22:13 +00:00
scoring_agent.py feat: add Go lem CLI and scoring-agent scripts 2026-02-15 16:22:13 +00:00
self_distill.py feat: LEK-1 kernel A/B test — 29 models, P100 validation, curriculum pipeline 2026-02-19 11:32:26 +00:00
sync_hf.py Add missing HF model cards, sync script, and Parquet export 2026-02-15 00:14:26 +00:00