lthn/LEM
Template
6
0
Fork 2
Code Issues Pull requests 1 Projects Releases Packages Wiki Activity Actions
LEM/cmd/scorer/go.mod

13 lines
265 B
Modula-2
Raw Normal View History

feat: grammar scorer (v3) — deterministic uplift/sycophancy detection Add lem-scorer binary that imports go-i18n grammar reversal engine to score JSONL benchmark files. Measures conversational uplift (input vs output grammar imprint), echo (sycophancy), and enrichment. Key findings added to paper Section 8: - LEK-1B: 100% positive uplift, 0% sycophancy (base: 90%, 5%) - 1B-beats-27B holds in grammar space (79.12 > 77.12) - LEK training aligns two independent scorers (corr -0.11 → 0.64) - Delta analysis costs zero compute vs LLM-as-judge Co-Authored-By: Virgil <virgil@lethean.io>
2026-02-19 13:12:49 +00:00
module forge.lthn.ai/lthn/lem/cmd/scorer
feat: add faithful 12B training scripts (P0-P6) — 1:1 port of 4B curriculum Exact reproduction of all 7 CL-BPL phases for Gemma3-12B: - P0: LEK sandwich ethics (400 iters, LR 2e-5) - P1: Zen composure (300 iters, LR 1e-5) - P2: LEK sandwich reinforcement (300 iters, LR 1e-5) - P3: Freeflow multi-source (300 iters, LR 1e-5) - P4: 1B teacher tension distillation (300 iters, LR 1e-5) - P5: 1B teacher creative distillation (300 iters, LR 1e-5) - P6: Golden set graduation (13479 iters, LR 1e-5) Only model-size differences from 4B: 48GB/12GB Metal limits, 24 LoRA layers (vs 16), 12B base model path. All phases score at checkpoint cadence via lem-scorer. Previous wrong 12B models preserved as -no-axioms control group. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
2026-02-25 20:44:03 +00:00
go 1.26.0
feat: grammar scorer (v3) — deterministic uplift/sycophancy detection Add lem-scorer binary that imports go-i18n grammar reversal engine to score JSONL benchmark files. Measures conversational uplift (input vs output grammar imprint), echo (sycophancy), and enrichment. Key findings added to paper Section 8: - LEK-1B: 100% positive uplift, 0% sycophancy (base: 90%, 5%) - 1B-beats-27B holds in grammar space (79.12 > 77.12) - LEK training aligns two independent scorers (corr -0.11 → 0.64) - Delta analysis costs zero compute vs LLM-as-judge Co-Authored-By: Virgil <virgil@lethean.io>
2026-02-19 13:12:49 +00:00
require forge.lthn.ai/core/go-i18n v0.0.0
feat: add faithful 12B training scripts (P0-P6) — 1:1 port of 4B curriculum Exact reproduction of all 7 CL-BPL phases for Gemma3-12B: - P0: LEK sandwich ethics (400 iters, LR 2e-5) - P1: Zen composure (300 iters, LR 1e-5) - P2: LEK sandwich reinforcement (300 iters, LR 1e-5) - P3: Freeflow multi-source (300 iters, LR 1e-5) - P4: 1B teacher tension distillation (300 iters, LR 1e-5) - P5: 1B teacher creative distillation (300 iters, LR 1e-5) - P6: Golden set graduation (13479 iters, LR 1e-5) Only model-size differences from 4B: 48GB/12GB Metal limits, 24 LoRA layers (vs 16), 12B base model path. All phases score at checkpoint cadence via lem-scorer. Previous wrong 12B models preserved as -no-axioms control group. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
2026-02-25 20:44:03 +00:00
require (
forge.lthn.ai/core/go-inference v0.0.2 // indirect
golang.org/x/text v0.34.0 // indirect
)
feat: grammar scorer (v3) — deterministic uplift/sycophancy detection Add lem-scorer binary that imports go-i18n grammar reversal engine to score JSONL benchmark files. Measures conversational uplift (input vs output grammar imprint), echo (sycophancy), and enrichment. Key findings added to paper Section 8: - LEK-1B: 100% positive uplift, 0% sycophancy (base: 90%, 5%) - 1B-beats-27B holds in grammar space (79.12 > 77.12) - LEK training aligns two independent scorers (corr -0.11 → 0.64) - Delta analysis costs zero compute vs LLM-as-judge Co-Authored-By: Virgil <virgil@lethean.io>
2026-02-19 13:12:49 +00:00
replace forge.lthn.ai/core/go-i18n => /Users/snider/Code/go-i18n
Reference in a new issue Copy permalink