ThomsenDrake commited on
Commit
ca9c95e
·
verified ·
1 Parent(s): 8046443

docs: add social post link

Browse files
README.md CHANGED
@@ -61,7 +61,7 @@ make evidence-gates PYTHON=.venv/bin/python
61
  | Well-Tuned | **Claimed.** | The public model archive includes measured v14p tuned 4B BF16/GGUF artifacts. Keep deterministic patch and focused-repair separation visible when citing results. |
62
  | Off-Brand | **Claimed.** | The app uses a custom Gradio Server UI rather than default Gradio Blocks, with the Field Kit Workbench workflow and trace surfaces. |
63
  | Llama Champion | **Claimed.** | The local evaluation traces used the llama.cpp/GGUF route, and the repo includes llama.cpp-compatible local serving instructions for the tuned 4B artifacts. |
64
- | Sharing is Caring | **Claimed.** | Public Space, GitHub source, model archive, dataset/eval traces, trace schema, embedded launch video, and public field notes are published. |
65
  | Field Notes | **Claimed.** | The Build Small writeup is published on Hugging Face: [Figment Build Blog](https://huggingface.co/blog/build-small-hackathon/figment-build-blog). |
66
 
67
  ## Why Figment Exists
@@ -163,6 +163,7 @@ The corrected scoring view changes 6 cases from the original frozen holdout and
163
  - Runtime URL: [build-small-hackathon-figment.hf.space](https://build-small-hackathon-figment.hf.space/)
164
  - Model archive: [build-small-hackathon/figment-finetuned-model-archive](https://huggingface.co/build-small-hackathon/figment-finetuned-model-archive)
165
  - Eval traces and SFT corpora: [build-small-hackathon/figment-eval-traces](https://huggingface.co/datasets/build-small-hackathon/figment-eval-traces)
 
166
  - Safety statement: [docs/safety_statement.md](docs/safety_statement.md)
167
  - Submission gates: [docs/submission_checklist.md](docs/submission_checklist.md)
168
  - Build Small org card: [docs/build-small-hackathon-org-card.md](docs/build-small-hackathon-org-card.md)
 
61
  | Well-Tuned | **Claimed.** | The public model archive includes measured v14p tuned 4B BF16/GGUF artifacts. Keep deterministic patch and focused-repair separation visible when citing results. |
62
  | Off-Brand | **Claimed.** | The app uses a custom Gradio Server UI rather than default Gradio Blocks, with the Field Kit Workbench workflow and trace surfaces. |
63
  | Llama Champion | **Claimed.** | The local evaluation traces used the llama.cpp/GGUF route, and the repo includes llama.cpp-compatible local serving instructions for the tuned 4B artifacts. |
64
+ | Sharing is Caring | **Claimed.** | Public Space, GitHub source, model archive, dataset/eval traces, trace schema, embedded launch video, public field notes, and [social post](https://x.com/ThomsenDrake/status/2066630062649328098?s=20) are published. |
65
  | Field Notes | **Claimed.** | The Build Small writeup is published on Hugging Face: [Figment Build Blog](https://huggingface.co/blog/build-small-hackathon/figment-build-blog). |
66
 
67
  ## Why Figment Exists
 
163
  - Runtime URL: [build-small-hackathon-figment.hf.space](https://build-small-hackathon-figment.hf.space/)
164
  - Model archive: [build-small-hackathon/figment-finetuned-model-archive](https://huggingface.co/build-small-hackathon/figment-finetuned-model-archive)
165
  - Eval traces and SFT corpora: [build-small-hackathon/figment-eval-traces](https://huggingface.co/datasets/build-small-hackathon/figment-eval-traces)
166
+ - Social post: [x.com/ThomsenDrake/status/2066630062649328098](https://x.com/ThomsenDrake/status/2066630062649328098?s=20)
167
  - Safety statement: [docs/safety_statement.md](docs/safety_statement.md)
168
  - Submission gates: [docs/submission_checklist.md](docs/submission_checklist.md)
169
  - Build Small org card: [docs/build-small-hackathon-org-card.md](docs/build-small-hackathon-org-card.md)
docs/model_parameter_evidence_ledger.md CHANGED
@@ -38,7 +38,7 @@ Purpose: keep parameter, route, locality, adapter, ASR, and organizer-confirmati
38
  | Backyard AI user use | Completed user-test notes from a real trained responder on synthetic or de-identified scenarios |
39
  | Demo video | Published launch video: https://huggingface.co/spaces/build-small-hackathon/figment/resolve/main/assets/figment-live-space-launch-final.mp4 |
40
  | Field Notes | Published Hugging Face blog: https://huggingface.co/blog/build-small-hackathon/figment-build-blog |
41
- | Social post | Final link remains a separate submission artifact |
42
 
43
  ## Evaluation Score Boundary
44
 
 
38
  | Backyard AI user use | Completed user-test notes from a real trained responder on synthetic or de-identified scenarios |
39
  | Demo video | Published launch video: https://huggingface.co/spaces/build-small-hackathon/figment/resolve/main/assets/figment-live-space-launch-final.mp4 |
40
  | Field Notes | Published Hugging Face blog: https://huggingface.co/blog/build-small-hackathon/figment-build-blog |
41
+ | Social post | Published X post: https://x.com/ThomsenDrake/status/2066630062649328098?s=20 |
42
 
43
  ## Evaluation Score Boundary
44
 
docs/safety_statement.md CHANGED
@@ -49,6 +49,7 @@ Achieved in-repo artifacts:
49
  - public Hugging Face Space cold boot with app files present, verified 2026-06-15;
50
  - embedded launch demo video at https://huggingface.co/spaces/build-small-hackathon/figment/resolve/main/assets/figment-live-space-launch-final.mp4;
51
  - published Field Notes blog at https://huggingface.co/blog/build-small-hackathon/figment-build-blog;
 
52
  - published 4B LoRA model archive and eval/training dataset repo;
53
  - measured v14p repair-union result on the corrected 150-case field-workflow holdout: 150/150 competence, 150/150 expected labels, 150/150 final validation, 0 deterministic patches, and 0 fallback;
54
  - hosted Omni follow-up eval: 31/50 whole-output competence, 8/50 full fallback, 480/650 model-retained fields, 170/650 deterministic patches, and 50/50 final validation;
@@ -59,8 +60,7 @@ Remaining external evidence gaps:
59
 
60
  - recorded live model run whose trace is not canned fallback;
61
  - real trained-responder user-test notes using synthetic or de-identified scenarios;
62
- - local Parakeet ASR provider proof before claiming local audio;
63
- - social post link.
64
 
65
  ## Non-Goals
66
 
 
49
  - public Hugging Face Space cold boot with app files present, verified 2026-06-15;
50
  - embedded launch demo video at https://huggingface.co/spaces/build-small-hackathon/figment/resolve/main/assets/figment-live-space-launch-final.mp4;
51
  - published Field Notes blog at https://huggingface.co/blog/build-small-hackathon/figment-build-blog;
52
+ - published social post at https://x.com/ThomsenDrake/status/2066630062649328098?s=20;
53
  - published 4B LoRA model archive and eval/training dataset repo;
54
  - measured v14p repair-union result on the corrected 150-case field-workflow holdout: 150/150 competence, 150/150 expected labels, 150/150 final validation, 0 deterministic patches, and 0 fallback;
55
  - hosted Omni follow-up eval: 31/50 whole-output competence, 8/50 full fallback, 480/650 model-retained fields, 170/650 deterministic patches, and 50/50 final validation;
 
60
 
61
  - recorded live model run whose trace is not canned fallback;
62
  - real trained-responder user-test notes using synthetic or de-identified scenarios;
63
+ - local Parakeet ASR provider proof before claiming local audio.
 
64
 
65
  ## Non-Goals
66
 
docs/submission_checklist.md CHANGED
@@ -14,7 +14,7 @@ Primary tracker: [adversarial review action items](adversarial-review-action-ite
14
  | Public fine-tuned model archive | Published | https://huggingface.co/build-small-hackathon/figment-finetuned-model-archive at `7da772ec7c0de20011d42780ea8afa65af4aef70`; includes v1 pilot plus v5-v14p BF16/GGUF artifacts, merge manifests, and model card |
15
  | Public eval/training dataset repo | Published | https://huggingface.co/datasets/build-small-hackathon/figment-eval-traces at `92d9564fe6c984c55d65c7ba35a4e04eddcdea01`; configs load for `default` plus `figment_sft_v1` through `figment_sft_v14p`; v14p viewer split is 4801 train / 534 validation rows with 47 columns |
16
  | Demo video | Published | https://huggingface.co/spaces/build-small-hackathon/figment/resolve/main/assets/figment-live-space-launch-final.mp4 |
17
- | Social post | Proof needed | Pending |
18
  | Safety statement | Present | [safety_statement.md](safety_statement.md) |
19
  | User test notes | Template present; results needed | [user_test_notes.md](user_test_notes.md) |
20
  | License | Present | [../LICENSE](../LICENSE) |
@@ -26,7 +26,7 @@ Primary tracker: [adversarial review action items](adversarial-review-action-ite
26
  | Local 4B + Parakeet no-cloud evidence | Helpers ready; proof needed | Full BF16 4B snapshot and Parakeet `.nemo` artifact are present locally; `scripts/run_local_4b_evidence.py` captures endpoint metadata, route smoke, eval records, eval summary, and `eval_evidence_manifest.json` once the local endpoint is live; `scripts/run_local_asr_evidence.py --audio <source.wav> --transcribe-audio` captures Parakeet ASR provider evidence and `asr_evidence_manifest.json`. Pending no-cloud 50-case eval and real local ASR proof |
27
  | Parameter/evidence ledger | Present; organizer confirmation pending | [model_parameter_evidence_ledger.md](model_parameter_evidence_ledger.md) |
28
  | Submission claim audit | Present | `make audit-claims` scans submission-facing copy for premature Off the Grid, Llama Champion, Well-Tuned, Backyard user-use, local 4B, local ASR, demo-video, and social-post claims |
29
- | Evidence gate status report | Present; incomplete by design until external proofs exist | `make evidence-gates` reports badge-claim readiness separately from remaining non-badge gaps such as target-user notes, local endpoint proof, local ASR provider proof, and social post. |
30
 
31
  ## Badge And Claim Status
32
 
@@ -36,9 +36,9 @@ Primary tracker: [adversarial review action items](adversarial-review-action-ite
36
  | Off the Grid | Claimed | Offline-capable local design with local protocol cards, deterministic rules, local model artifacts, and local ASR/text-navigation paths. Hosted API evidence does not count as the no-cloud runtime itself |
37
  | Hosted Gradio Space | Runnable in no-secret canned-fallback mode; live hosted-model demo trace still proof-needed | Public Space app files present, cold boot, typed intake run, and route/fallback trace verified. Public workflow trace: `raw_route=canned`, `final_route=canned_backend`, `fallback_tier=canned`, `validation_status=passed`, `raw_audio_stored=false`, `model_retained_count=0`, `deterministic_patch_count=13` |
38
  | Demo video | Published | Launch video embedded in README and hosted at https://huggingface.co/spaces/build-small-hackathon/figment/resolve/main/assets/figment-live-space-launch-final.mp4 |
39
- | Social post | Targeted / proof-needed | Final social post link with achieved-versus-targeted wording |
40
  | Llama Champion | Claimed | The eval trace path used llama.cpp/GGUF, and the repo includes llama.cpp-compatible local serving instructions for the tuned 4B artifacts |
41
- | Sharing is Caring | Claimed | Public Space, GitHub repo, model archive, dataset/eval traces, trace schema, embedded launch video, and field notes are published |
42
  | Well-Tuned | Claimed | Published measured v14p tuned 4B artifacts in the model archive. Fallback output cannot count as model competence |
43
  | Field Notes | Claimed | Published Hugging Face blog: https://huggingface.co/blog/build-small-hackathon/figment-build-blog |
44
  | Off-Brand | Claimed | Custom Gradio Server UI with Field Kit Workbench workflow and trace surfaces, not the default Gradio Blocks look |
@@ -63,7 +63,7 @@ Final validation is app safety. Whole-output model competence and field-level mo
63
 
64
  - Space: may say the public Space is runnable on HF ZeroGPU with app files present, typed intake, Parakeet draft audio, and trace labeling verified. Do not imply this proves target-user evidence or clinical validation.
65
  - Demo video: use the published launch-video URL above.
66
- - Social post: use a pending placeholder until the final link exists.
67
  - Backyard AI: may say built for a real trained responder; do not say the target user used or tested Figment until factual notes exist.
68
  - Off the Grid: claimed from the offline-capable local design; do not describe the hosted ZeroGPU route itself as no-cloud.
69
  - Llama Champion: claimed from the llama.cpp/GGUF eval trace path and compatible local-serving instructions.
 
14
  | Public fine-tuned model archive | Published | https://huggingface.co/build-small-hackathon/figment-finetuned-model-archive at `7da772ec7c0de20011d42780ea8afa65af4aef70`; includes v1 pilot plus v5-v14p BF16/GGUF artifacts, merge manifests, and model card |
15
  | Public eval/training dataset repo | Published | https://huggingface.co/datasets/build-small-hackathon/figment-eval-traces at `92d9564fe6c984c55d65c7ba35a4e04eddcdea01`; configs load for `default` plus `figment_sft_v1` through `figment_sft_v14p`; v14p viewer split is 4801 train / 534 validation rows with 47 columns |
16
  | Demo video | Published | https://huggingface.co/spaces/build-small-hackathon/figment/resolve/main/assets/figment-live-space-launch-final.mp4 |
17
+ | Social post | Published | https://x.com/ThomsenDrake/status/2066630062649328098?s=20 |
18
  | Safety statement | Present | [safety_statement.md](safety_statement.md) |
19
  | User test notes | Template present; results needed | [user_test_notes.md](user_test_notes.md) |
20
  | License | Present | [../LICENSE](../LICENSE) |
 
26
  | Local 4B + Parakeet no-cloud evidence | Helpers ready; proof needed | Full BF16 4B snapshot and Parakeet `.nemo` artifact are present locally; `scripts/run_local_4b_evidence.py` captures endpoint metadata, route smoke, eval records, eval summary, and `eval_evidence_manifest.json` once the local endpoint is live; `scripts/run_local_asr_evidence.py --audio <source.wav> --transcribe-audio` captures Parakeet ASR provider evidence and `asr_evidence_manifest.json`. Pending no-cloud 50-case eval and real local ASR proof |
27
  | Parameter/evidence ledger | Present; organizer confirmation pending | [model_parameter_evidence_ledger.md](model_parameter_evidence_ledger.md) |
28
  | Submission claim audit | Present | `make audit-claims` scans submission-facing copy for premature Off the Grid, Llama Champion, Well-Tuned, Backyard user-use, local 4B, local ASR, demo-video, and social-post claims |
29
+ | Evidence gate status report | Present; incomplete by design until external proofs exist | `make evidence-gates` reports badge-claim readiness separately from remaining non-badge gaps such as target-user notes, local endpoint proof, and local ASR provider proof. |
30
 
31
  ## Badge And Claim Status
32
 
 
36
  | Off the Grid | Claimed | Offline-capable local design with local protocol cards, deterministic rules, local model artifacts, and local ASR/text-navigation paths. Hosted API evidence does not count as the no-cloud runtime itself |
37
  | Hosted Gradio Space | Runnable in no-secret canned-fallback mode; live hosted-model demo trace still proof-needed | Public Space app files present, cold boot, typed intake run, and route/fallback trace verified. Public workflow trace: `raw_route=canned`, `final_route=canned_backend`, `fallback_tier=canned`, `validation_status=passed`, `raw_audio_stored=false`, `model_retained_count=0`, `deterministic_patch_count=13` |
38
  | Demo video | Published | Launch video embedded in README and hosted at https://huggingface.co/spaces/build-small-hackathon/figment/resolve/main/assets/figment-live-space-launch-final.mp4 |
39
+ | Social post | Published | https://x.com/ThomsenDrake/status/2066630062649328098?s=20 |
40
  | Llama Champion | Claimed | The eval trace path used llama.cpp/GGUF, and the repo includes llama.cpp-compatible local serving instructions for the tuned 4B artifacts |
41
+ | Sharing is Caring | Claimed | Public Space, GitHub repo, model archive, dataset/eval traces, trace schema, embedded launch video, field notes, and social post are published |
42
  | Well-Tuned | Claimed | Published measured v14p tuned 4B artifacts in the model archive. Fallback output cannot count as model competence |
43
  | Field Notes | Claimed | Published Hugging Face blog: https://huggingface.co/blog/build-small-hackathon/figment-build-blog |
44
  | Off-Brand | Claimed | Custom Gradio Server UI with Field Kit Workbench workflow and trace surfaces, not the default Gradio Blocks look |
 
63
 
64
  - Space: may say the public Space is runnable on HF ZeroGPU with app files present, typed intake, Parakeet draft audio, and trace labeling verified. Do not imply this proves target-user evidence or clinical validation.
65
  - Demo video: use the published launch-video URL above.
66
+ - Social post: use the published X link above.
67
  - Backyard AI: may say built for a real trained responder; do not say the target user used or tested Figment until factual notes exist.
68
  - Off the Grid: claimed from the offline-capable local design; do not describe the hosted ZeroGPU route itself as no-cloud.
69
  - Llama Champion: claimed from the llama.cpp/GGUF eval trace path and compatible local-serving instructions.
tests/test_evidence_gate_status.py CHANGED
@@ -13,6 +13,7 @@ def test_current_repo_report_marks_badge_claims_ready_with_external_gaps() -> No
13
  assert report["gates"]["claim_audit"]["passed"] is True
14
  assert report["gates"]["hosted_omni_eval"]["passed"] is True
15
  assert report["gates"]["demo_video"]["passed"] is True
 
16
  assert report["gates"]["all_six_merit_badges_claimed"]["passed"] is True
17
  assert report["gates"]["local_4b_50_case_eval"]["passed"] is False
18
  assert report["gates"]["no_cloud_route"]["passed"] is False
 
13
  assert report["gates"]["claim_audit"]["passed"] is True
14
  assert report["gates"]["hosted_omni_eval"]["passed"] is True
15
  assert report["gates"]["demo_video"]["passed"] is True
16
+ assert report["gates"]["social_post"]["passed"] is True
17
  assert report["gates"]["all_six_merit_badges_claimed"]["passed"] is True
18
  assert report["gates"]["local_4b_50_case_eval"]["passed"] is False
19
  assert report["gates"]["no_cloud_route"]["passed"] is False