File size: 3,318 Bytes
294f8ea
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
7364912
 
 
 
 
294f8ea
 
 
 
 
 
7364912
 
 
 
294f8ea
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
# Publication guide

OpenSysOne is an independent decision-scoring experiment inspired by
[Jev](https://typesafe.ai/) and the TypeSafe team. The current release is the
completed, calibrated Qwen3 4B scorer evaluated on 17 September 2026.

The published repository has a short entry path:

| Directory | Contents |
| --- | --- |
| [model/](../model/README.md) | Calibrated checkpoint, exact hash, base-model requirements and reconstruction constraints |
| [results/](../results/) | [Final report](../results/report.md), metrics, CSV tables and standalone charts |
| [docs/](README.md) | This guide and [reproduction instructions](reproduce.md) |
| [source/](../source/) | Complete committed project tree, including code, tests, examples, frontend, documentation and small evidence |
| [archive/](../archive/README.md) | Index to historical checkpoints, source revisions and publication records |

The [API guide](../source/docs/usage/jev-api.md) describes the Jev-compatible
request shape. The [playground guide](../source/docs/usage/playground.md) describes
the local browser interface. Neither the Hugging Face repository nor its model
card is a hosted inference service.

## Current files and immutable history

The front directories provide convenient copies and navigation. Exact release
identity comes from the existing pointers and their recorded Hub payload commits:

- [FINAL_MODEL.json](../FINAL_MODEL.json): calibrated model hash and original final-evaluation manifest.
- [PROFILE_RESULTS.json](../PROFILE_RESULTS.json): completed profiling, stopped-training evidence and source archives.
- [CURRENT_SNAPSHOT.json](../CURRENT_SNAPSHOT.json): earlier training snapshot, including resumable state; it is not the final-model pointer.

Historical `final/`, `profiles/`, `snapshots/`, `sources/` and `publications/`
payloads remain available at their recorded paths. Their manifests and checksums
are not rewritten to fit this presentation. Use a pointer's `payload_commit`
when retrieving its `path` and `manifest_path` for a reproducible download.

`source/` is the complete publication source tree. The exact training and
evaluation revisions are separately recorded in artifact metadata and immutable
source archives; a later documentation revision is not a new model training run.
Keep source-relative paths intact when executing commands.

The subsequent Apache-2.0 license adoption is recorded in
[`LICENSING.json`](../LICENSING.json). `PUBLICATION.json` continues to identify
the earlier verified presentation at its recorded payload commit; its manifest
and the historical source archives are preserved.

## Scope

The calibrated checkpoint contains adapter/head parameters and requires the
pinned pretrained base. It is a custom OpenSysOne artifact. Results support the
reported benchmark comparisons, with separate calibration and validation-only
selection; they do not establish general intelligence or calibration on arbitrary
tasks. Original OpenSysOne code, documentation, and adapter/head additions are
licensed under [Apache License 2.0](../LICENSE); see [NOTICE](../NOTICE) for
attribution and scope. Upstream model and dataset terms and notices remain
applicable. See the [model notes](../model/README.md) and
[measured report](../results/report.md) before interpreting probabilities or speed.