Lory-4B v1

The small, cheap tier of Lory v1: a decision model for simulated survey respondents that responds to the stakes in a question (prices, odds, payoffs, delays), returning a calibrated probability for every option in one forward pass. A LoRA adapter + pointer head on Kev-4B (Qwen3.5-4B-Base), served with kev.serve (System One API). v1 (public preview): one training seed. For the strongest economic logic use Lory-27B; this model is the budget option.

What changed

Same EDSL survey, two respondents (a budget-conscious teacher on $52k, a surgeon on $480k):

question	released Kev-4B	this model
Streaming service at $8 vs $40/month, teacher: P(subscribe)	0.49 → 0.25	0.88 → 0.05
same, surgeon	0.78 → 0.77	0.96 → 0.51
Picks the dominated plan ($45 for 5 GB next to $30 for 5 GB)	0.24–0.29	0.02–0.03
Surgeon "very likely" to buy an air fryer at $8 vs $40	0.25 → 0.35 (wrong way)	0.82 → 0.67

Results

	DLS v0.2 (ours, 9 families)	blind test (513 contrasts by GPT-6.1-sol)	bank-run battery: payoff shift (theory 80)	lottery slope (theory 1.0)	EDSL-General
this model	79 [73, 86]	81 [78, 84]	28	0.51	85
Kev-4B (released)	53	67	−1	0.05	85
Lory-27B v1	88	92	66	0.77	88
Perplexity Decider v1.2	77	93	9	0.43	88

Gains over the released model are large and robust (DLS +25.5 [+17.9, +32.9], blind +14.0 [+10.4, +17.5]). It ties Decider on DLS (+1.6 [−5.2, +8.2]) but is clearly behind it on the blind test (−12.7 [−15.5, −10.1]). Test descriptions: see the Lory-27B card.

Use

git clone https://github.com/jaredpalmer/kev && cd kev && git checkout 2074cc4 && uv sync --extra serve
uv run python -m kev.serve --run eltokh7/lory-4b --device cuda --port 8850     # or --device mps on Apple silicon

Requests as in the Lory-27B card; EDSL surveys via edsl_decision.py (in this repo). Cost: at least 53 requests/s on one A100 (client-limited), under $0.25 per 1,000 respondents × 20 questions.

Training

The same 27,000 records and recipe as Lory-27B (LoRA delta from the released adapter, + 5,000 Kev decision-v7 replay, one epoch, lr 2e-5, batch 8, bf16), one A100, ~1.8 hours.

Limitations

Everything on the Lory-27B card, plus: it carries the logic to unfamiliar wording much less well than the 27B (bank-run payoff shift 28 vs 66); the persona signal is slightly weaker than before training on the blind test (69 vs 72); framing invariance 56/100.

Attribution

Kev and Kev-4B by Jared Palmer (Apache-2.0); backbone Qwen3.5-4B-Base (Qwen, Apache-2.0).