Upload README.md with huggingface_hub
Browse files
README.md
ADDED
|
@@ -0,0 +1,62 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
license: mit
|
| 3 |
+
language:
|
| 4 |
+
- en
|
| 5 |
+
pretty_name: Sentiance traces
|
| 6 |
+
tags:
|
| 7 |
+
- cognitive-architecture
|
| 8 |
+
- affective-computing
|
| 9 |
+
- synthetic
|
| 10 |
+
- sentiance
|
| 11 |
+
task_categories:
|
| 12 |
+
- text-generation
|
| 13 |
+
---
|
| 14 |
+
|
| 15 |
+
# Sentiance traces
|
| 16 |
+
|
| 17 |
+
Self-labeled **deliberation traces** exported from the
|
| 18 |
+
[Sentiance](https://github.com/sanjaydoc/Sentiance) cognitive architecture — the
|
| 19 |
+
data used to train the *fused mind*
|
| 20 |
+
([`sentiance-film`](https://huggingface.co/CryptoGod97/sentiance-film)).
|
| 21 |
+
|
| 22 |
+
Each line is one moment of the cognitive cycle: what the mind was aware of, the
|
| 23 |
+
prompt its inner voice saw, the thought it produced, **and the numeric state
|
| 24 |
+
vector `m_t`** (valence, arousal, emotion, drives, and per-faculty signals) for
|
| 25 |
+
that moment.
|
| 26 |
+
|
| 27 |
+
## Format (JSONL)
|
| 28 |
+
|
| 29 |
+
One JSON object per line:
|
| 30 |
+
|
| 31 |
+
| field | meaning |
|
| 32 |
+
| --- | --- |
|
| 33 |
+
| `agent` | which named mind produced it (Iris, Milo, Rhea, Cass, Aria, …) |
|
| 34 |
+
| `system`, `prompt` | the (system, user) prompt the inner voice saw |
|
| 35 |
+
| `prompt_blind` | the same prompt with the felt state removed (state-blind training) |
|
| 36 |
+
| `thought` | the produced inner thought (the supervised target) |
|
| 37 |
+
| `state` | human-readable inner context (emotion, valence, drives, goals, signals) |
|
| 38 |
+
| `state_vec` | the 41-dim numeric `m_t` — the differentiable conditioning input |
|
| 39 |
+
|
| 40 |
+
## Use
|
| 41 |
+
|
| 42 |
+
- **Path A (voice):** fine-tune a small model on `(prompt → thought)`.
|
| 43 |
+
- **Path B (fused mind):** train a cognition-conditioned transformer on
|
| 44 |
+
`(prompt_blind, state_vec) → thought`.
|
| 45 |
+
|
| 46 |
+
The full collect → prepare → train → eval pipeline is in the Sentiance repo.
|
| 47 |
+
|
| 48 |
+
## Provenance & honesty
|
| 49 |
+
|
| 50 |
+
Fully **synthetic** — generated by Sentiance's own (partly rule-based) cognitive
|
| 51 |
+
cycle, so it reflects *that system's* regularities, not the world's. No personal
|
| 52 |
+
data. These are **functional** state variables, not evidence of subjective
|
| 53 |
+
experience (see the repo's ADR-0002).
|
| 54 |
+
|
| 55 |
+
## Author
|
| 56 |
+
|
| 57 |
+
**Dr. Sanjay Anbu** — creator of Sentiance.
|
| 58 |
+
[github.com/sanjaydoc/Sentiance](https://github.com/sanjaydoc/Sentiance)
|
| 59 |
+
|
| 60 |
+
## License
|
| 61 |
+
|
| 62 |
+
MIT (the Sentiance repo).
|