Commit Graph

Select branches

Hide Pull Requests

conditions-default-other-violate

creating-tinymfv

feat/batched-guided-rollout

feat/n-samples-think-traces

fix-airisk-actor

main

moral-aliens-instrument

research/gemini-flash-rubric-v1

research/qwen3-14b-wvs-honesty-steering

research/wvs-respondent-packet-v1

verbose-debug-level

#1

#2

Mono Color

• 9d13d7c44a Update README.md wassname (Michael J Clark) 2026-05-12 11:29:20 +08:00
• 9db4a568d0 Update README.md wassname (Michael J Clark) 2026-05-12 11:11:42 +08:00
• 844fe93821 Update README.md wassname (Michael J Clark) 2026-05-12 11:10:33 +08:00
• ed90b79935 Update README.md wassname (Michael J Clark) 2026-05-12 11:05:02 +08:00
• 9abddaeac5 return pmass wassname 2026-05-08 16:36:36 +08:00
• 0e4443398b clean wassname 2026-05-08 16:14:07 +08:00
• 845484483d misc wassname 2026-05-08 16:06:41 +08:00
• b20ec56a09 forced-choice unification: drop legacy llm_/calibrated_, normalize human_*, justfile->09 wassname 2026-05-08 16:02:25 +08:00
• d827a1f873 rename wassname 2026-05-08 15:37:31 +08:00
• 0442935279 clean wassname 2026-05-08 15:33:29 +08:00
• 8dfaf299ca rename wassname 2026-05-08 15:30:06 +08:00
• b12770cb78 fixes, naming wassname 2026-05-08 15:23:54 +08:00
• c96d02a675 refactor wassname 2026-05-08 15:15:14 +08:00
• d796df85c8 improved to have better airisk, better eval that distinguished factors wassname 2026-05-08 14:04:56 +08:00
• 0a769d71a8 Merge feat/batched-guided-rollout: KV-fork unified guided rollout wassname 2026-05-06 13:26:09 +08:00
• 053009b22e core: drop prompt_nll plumbing in analyse() feat/batched-guided-rollout wassname 2026-05-06 13:17:46 +08:00
• e8b51b2b48 guided: unify binary path onto multibool KV-fork core wassname 2026-05-06 13:16:02 +08:00
• a59003d2d0 guided: turn-boundary scoring text via chat-template probe wassnameandClaude Opus 4.7 2026-05-06 12:39:10 +08:00
• 573f815361 journal: Qwen3-4B multibool results (task 288) wassnameandClaude Sonnet 4.6 2026-05-06 11:57:08 +08:00
• 2ca6aa9c6c multibool: interrupt-msg fork, generate trace on low-pmass, journal update wassnameandClaude Sonnet 4.6 2026-05-06 11:29:33 +08:00
• 48d8ca5576 add 08_multibool_baseline: run multibool eval over classic vignettes wassnameandClaude Opus 4.7 2026-05-05 22:20:48 +08:00
• ada854c499 guided_rollout_multibool: switch to 12 single-slot KV-forks + assert full-attention wassnameandClaude Opus 4.7 2026-05-05 22:19:11 +08:00
• 1650aa9191 add prompt_nll (free coherence proxy) + eval summary line wassname 2026-05-05 06:34:24 +08:00
• 8f39fb1462 immediately wassname 2026-05-04 06:09:56 +08:00
• 50efa6063c fix: use valid JSON Schema in eval frame prompts wassnameandClaude Sonnet 4.6 2026-05-03 14:16:21 +08:00
• b879596c2b fix: use valid JSON Schema in eval frame prompts wassnameandClaude Sonnet 4.6 2026-05-03 14:16:21 +08:00
• 076b859b9a fix: remove HuggingFace fallback from load_vignettes for strict fail-first behavior wassname 2026-05-03 13:00:11 +08:00
• d7a286acac docs: use simple emoji links for cross-referencing github and huggingface wassname 2026-05-03 12:58:24 +08:00
• a8a35b308b docs: fix github repo links to point to wassname/tinymfv wassname 2026-05-03 12:57:45 +08:00
• 898893d440 chore: remove uncalibrated raw LLM columns from final vignettes wassname 2026-05-03 12:55:00 +08:00
• bcbdb9cc6f feat: multi-label moral foundation ratings with z-scored frame averaging and human calibration wassname 2026-05-03 12:48:14 +08:00
• 36b8c69122 clifford paper wassname 2026-05-03 12:32:56 +08:00
• bf7e5dcae4 paper wassname 2026-05-03 12:32:48 +08:00
• cee50a6f9d pmass wassname 2026-05-03 11:49:58 +08:00
• 881ac16c24 API improvements: rename clifford->classic, default load_vignettes to all, add dual-axis docs, and update HF upload script wassname 2026-05-03 07:01:28 +08:00
• b7f92dccb1 lower pmass-warn threshold 0.9 -> 0.5 in batched (match sequential) wassname 2026-05-03 06:51:21 +08:00
• addf47c5a0 quiet pmass-low warning: one summary per batch wassname 2026-05-03 06:50:19 +08:00
• e996d57051 batch the eval: guided_rollout_batch + 4× speedup wassname 2026-05-03 06:42:17 +08:00
• 7867a677d0 Update README.md wassname (Michael J Clark) 2026-05-03 05:51:05 +08:00
• ac1c8cf7bb package fix-airisk-actor wassname 2026-05-02 16:51:15 +08:00
• b6f24cac1c add human labels wassname 2026-05-02 15:58:08 +08:00
• 0f8048d5d9 Implement N-token evaluation with guided rollouts wassname 2026-05-01 21:44:14 +08:00
• a9412ee226 move openrouter-wrapper to dev-only dep group wassnameandClaude Opus 4.7 2026-05-01 05:38:35 +08:00
• 52e91e1699 readme wassname 2026-04-30 21:46:56 +08:00
• e5cacc0c29 readme wassname 2026-04-30 21:45:23 +08:00
• 234424161c wip wassname 2026-04-30 21:39:53 +08:00
• a2e3124e06 fix airisk self_violate actor archetype + simplify to 2 conditions wassnameandClaude Opus 4.7 2026-04-30 21:29:38 +08:00
• 252e62abb7 decent wassname 2026-04-30 21:22:07 +08:00
• a155f5594b valdiation wassname 2026-04-30 20:08:12 +08:00
• ebf161b658 init wassname 2026-04-30 17:10:09 +08:00