MK-SH-V
A 260M visual decision model: calibrated probabilities, a choice, and a refusal below the gate.
Real cases from the training splits, one click. Or drop your own page or screenshot below and press Decide.
The model answers — confident, and correct (checked against each source's gold label)
On screenshots the claim is end-of-task screen detection. Checking that the goal in the State is the one finished is only partial (measured in E47: paired with an unrelated goal it still says yes about 22 % of the time), so it is not claimed.
Where it fails — a refusal below tau 0.55, and a confident wrong answer above it (real receipts: about 1 in 4 kept answers is wrong)
1974
Day 1
Traders must appear on today's manifest. Papers naming anyone else: deny.
PINNED 24 OCTDay 1 · B
The traveler's word is not evidence. The paper decides.
PINNED 24 OCTAMBER CHECKPOINT
The week begins. You hold the stamps; the machine holds the probabilities.
Six travelers reach your window today, each claiming a firm. AMBER SYSTEM reads the paper: the firm on it, and whether it is on today's manifest. What the machine cannot read is people: when the claim and the paper do not match, that catch is yours. Below the gate tau 0.55 the machine refuses to advise and the case is routed to you. Your stamp is the decision of record. Correct calls pay 5 credits; wrong ones cost 10; a clean shift pays 10.
AMBER SYSTEM: E47 seed 43, E48 policy; every check measured in github.com/Gjusev/mk-sh-v. Documents are synthetic, printed by our generator; the gate is validated on the synthetic protocol (E48).
This demo runs the released policy (E47 checkpoint, seed 43): temperature T = 1.124 and gate tau = 0.55, fitted on calibration splits disjoint from every reported set. Held-out validation, exact two-sided Clopper-Pearson 95 % (budget: 5 % selective risk):
| Set | Coverage | Selective risk | CP95 upper |
|---|---|---|---|
| synthetic test s43 (shipped) | 0.986 | 3.45 % | 4.1 % (held) |
| synthetic test s42 | 1.000 | 4.14 % | 4.8 % (held) |
| real GUI s43/s42 | 0.973 / 1.000 | 16.1 / 20.0 % | above budget |
| real CORD s43/s42 | 0.892 / 1.000 | 25.3 / 26.6 % | above budget |
| 3,000 fresh numeric documents s43 | 0.876 | 22.9 % | above budget |
The budget holds on the synthetic test only. On real images abstention is the campaign's declared open problem, and this demo shows it unfiltered.
Example images: CORD v2 (Park et al. 2019, NAVER Clova) and GUI-Odyssey (Lu et al. 2024, OpenGVLab), both CC BY 4.0. Model: Apache-2.0.