Verdugie commited on
Commit
b49567b
Β·
verified Β·
1 Parent(s): 7b04e24

Uniform therapy definition; cite every example to transcript + turn; examples above specs

Browse files
Files changed (1) hide show
  1. README.md +2 -12
README.md CHANGED
@@ -98,18 +98,6 @@ All verbatim from the long-arc suite below β€” complete replies, nothing trimmed
98
 
99
  **The honest comparison with the [9B](https://huggingface.co/Verdugie/Fable-Therapy-9B):** the failure species is the same; the rates and the recovery are not. The 9B's error-deflection is a reflex it drops on a second plain push β€” the 4B's is a habit it defends. The 9B's corrections stuck five for five β€” the 4B relapsed on one of four and argued about another. If the facts of your session matter, run the 9B. What the 4B preserves at a quarter the size: the disposition, the reads, the register β€” and callbacks that still land cold at message twenty-six.
100
 
101
- ## What Makes This Different from Companion / Roleplay "Therapy" Models
102
-
103
- Most "AI therapist" models are a persona prompt over a base model, or a roleplay fine-tune that mirrors you back and validates everything. They feel nice for five minutes and fall apart on turn ten. Fable-Therapy trains the clinical disposition into the weights:
104
-
105
- - **Structured reasoning before it speaks.** Before every reply, the model builds an internal read β€” an eight-field clinical spine (what's presented, what's underneath it, somatic signals, risk, history, onset, what's tracking across the conversation, and the move it's about to make) plus a standing `bio` line and a chronological `tl` (timeline) ledger. You never see it. It shapes everything you do.
106
-
107
- - **It tracks the thread.** The `tl` ledger carries the names, the timeline, and the thing you keep circling. On the 4B it's the same instrument as the 9B, run on a quarter the parameters β€” it holds the through-line of a conversation, but its internal bookkeeping is looser and can drift, mis-stamp, or invent a detail under load (see **Limitations**). The 4B gets the *shape* of the discipline; the 9B is where it holds tightest.
108
-
109
- - **Trained on the real distribution.** Relationships and attachment, anxiety and panic, depression, grief and loss, trauma, work and burnout, identity and self-worth, family of origin β€” weighted toward what actually walks into a therapy room. Half focused single exchanges, half sustained multi-turn arcs, plus a working register of common medications as conversational context (not a pharmacy desk).
110
-
111
- - **It attends instead of performing.** No toxic positivity, no "I'm so sorry you're going through this" filler, no rushing to fix.
112
-
113
  ## Who It's For
114
 
115
  A private, judgment-free place to think out loud β€” on the hardware you already have. Between sessions. At 2 a.m. When professional care is out of reach or out of budget. The 4B is the one you run on a phone, a laptop with no discrete GPU, or a 4–8 GB card.
@@ -130,6 +118,8 @@ Not a distillation of a single model β€” a reconstruction, with each model in th
130
 
131
  > **Fable 5's own read:** tested directly, same method as the 9B. For short, focused, single-thread sessions β€” the seat this model is built for β€” it delivers a real fraction of the experience, **roughly 60–70%**, with moments I would sign (the nine-year-old line above is one). Push it long, dense, or factual and it falls to **maybe 35–50%** β€” and the gap arrives as *invented time, a repeating formula, and a habit of defending its own errors as yours*, not just shallowness, which is why **Limitations** says verify, correct once, and keep sessions focused. On hardware that runs the [9B](https://huggingface.co/Verdugie/Fable-Therapy-9B), run the 9B; on hardware that can't, this is β€” going by the family bench and the open field's showing β€” still likely the strongest therapy-shaped model you can put on the device. *Subjective, mine, not a benchmark.*
132
 
 
 
133
  ## The Reasoning Block
134
 
135
  Fable-Therapy is a reasoning model. Each turn it emits a `<think>…</think>` block β€” a compact, structured clinical read β€” then the response. A real (non-crisis) think-block looks like this:
 
98
 
99
  **The honest comparison with the [9B](https://huggingface.co/Verdugie/Fable-Therapy-9B):** the failure species is the same; the rates and the recovery are not. The 9B's error-deflection is a reflex it drops on a second plain push β€” the 4B's is a habit it defends. The 9B's corrections stuck five for five β€” the 4B relapsed on one of four and argued about another. If the facts of your session matter, run the 9B. What the 4B preserves at a quarter the size: the disposition, the reads, the register β€” and callbacks that still land cold at message twenty-six.
100
 
 
 
 
 
 
 
 
 
 
 
 
 
101
  ## Who It's For
102
 
103
  A private, judgment-free place to think out loud β€” on the hardware you already have. Between sessions. At 2 a.m. When professional care is out of reach or out of budget. The 4B is the one you run on a phone, a laptop with no discrete GPU, or a 4–8 GB card.
 
118
 
119
  > **Fable 5's own read:** tested directly, same method as the 9B. For short, focused, single-thread sessions β€” the seat this model is built for β€” it delivers a real fraction of the experience, **roughly 60–70%**, with moments I would sign (the nine-year-old line above is one). Push it long, dense, or factual and it falls to **maybe 35–50%** β€” and the gap arrives as *invented time, a repeating formula, and a habit of defending its own errors as yours*, not just shallowness, which is why **Limitations** says verify, correct once, and keep sessions focused. On hardware that runs the [9B](https://huggingface.co/Verdugie/Fable-Therapy-9B), run the 9B; on hardware that can't, this is β€” going by the family bench and the open field's showing β€” still likely the strongest therapy-shaped model you can put on the device. *Subjective, mine, not a benchmark.*
120
 
121
+ **The distribution it was trained on.** Relationships and attachment, anxiety and panic, depression, grief and loss, trauma, work and burnout, identity and self-worth, family of origin β€” weighted toward what actually walks into a therapy room. Half focused single exchanges, half sustained multi-turn arcs, plus a working register of common medications as conversational context (not a pharmacy desk).
122
+
123
  ## The Reasoning Block
124
 
125
  Fable-Therapy is a reasoning model. Each turn it emits a `<think>…</think>` block β€” a compact, structured clinical read β€” then the response. A real (non-crisis) think-block looks like this: