agent-comparability 0.5.2__tar.gz

This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
@@ -0,0 +1,202 @@
1
+
2
+ Apache License
3
+ Version 2.0, January 2004
4
+ http://www.apache.org/licenses/
5
+
6
+ TERMS AND CONDITIONS FOR USE, REPRODUCTION, AND DISTRIBUTION
7
+
8
+ 1. Definitions.
9
+
10
+ "License" shall mean the terms and conditions for use, reproduction,
11
+ and distribution as defined by Sections 1 through 9 of this document.
12
+
13
+ "Licensor" shall mean the copyright owner or entity authorized by
14
+ the copyright owner that is granting the License.
15
+
16
+ "Legal Entity" shall mean the union of the acting entity and all
17
+ other entities that control, are controlled by, or are under common
18
+ control with that entity. For the purposes of this definition,
19
+ "control" means (i) the power, direct or indirect, to cause the
20
+ direction or management of such entity, whether by contract or
21
+ otherwise, or (ii) ownership of fifty percent (50%) or more of the
22
+ outstanding shares, or (iii) beneficial ownership of such entity.
23
+
24
+ "You" (or "Your") shall mean an individual or Legal Entity
25
+ exercising permissions granted by this License.
26
+
27
+ "Source" form shall mean the preferred form for making modifications,
28
+ including but not limited to software source code, documentation
29
+ source, and configuration files.
30
+
31
+ "Object" form shall mean any form resulting from mechanical
32
+ transformation or translation of a Source form, including but
33
+ not limited to compiled object code, generated documentation,
34
+ and conversions to other media types.
35
+
36
+ "Work" shall mean the work of authorship, whether in Source or
37
+ Object form, made available under the License, as indicated by a
38
+ copyright notice that is included in or attached to the work
39
+ (an example is provided in the Appendix below).
40
+
41
+ "Derivative Works" shall mean any work, whether in Source or Object
42
+ form, that is based on (or derived from) the Work and for which the
43
+ editorial revisions, annotations, elaborations, or other modifications
44
+ represent, as a whole, an original work of authorship. For the purposes
45
+ of this License, Derivative Works shall not include works that remain
46
+ separable from, or merely link (or bind by name) to the interfaces of,
47
+ the Work and Derivative Works thereof.
48
+
49
+ "Contribution" shall mean any work of authorship, including
50
+ the original version of the Work and any modifications or additions
51
+ to that Work or Derivative Works thereof, that is intentionally
52
+ submitted to Licensor for inclusion in the Work by the copyright owner
53
+ or by an individual or Legal Entity authorized to submit on behalf of
54
+ the copyright owner. For the purposes of this definition, "submitted"
55
+ means any form of electronic, verbal, or written communication sent
56
+ to the Licensor or its representatives, including but not limited to
57
+ communication on electronic mailing lists, source code control systems,
58
+ and issue tracking systems that are managed by, or on behalf of, the
59
+ Licensor for the purpose of discussing and improving the Work, but
60
+ excluding communication that is conspicuously marked or otherwise
61
+ designated in writing by the copyright owner as "Not a Contribution."
62
+
63
+ "Contributor" shall mean Licensor and any individual or Legal Entity
64
+ on behalf of whom a Contribution has been received by Licensor and
65
+ subsequently incorporated within the Work.
66
+
67
+ 2. Grant of Copyright License. Subject to the terms and conditions of
68
+ this License, each Contributor hereby grants to You a perpetual,
69
+ worldwide, non-exclusive, no-charge, royalty-free, irrevocable
70
+ copyright license to reproduce, prepare Derivative Works of,
71
+ publicly display, publicly perform, sublicense, and distribute the
72
+ Work and such Derivative Works in Source or Object form.
73
+
74
+ 3. Grant of Patent License. Subject to the terms and conditions of
75
+ this License, each Contributor hereby grants to You a perpetual,
76
+ worldwide, non-exclusive, no-charge, royalty-free, irrevocable
77
+ (except as stated in this section) patent license to make, have made,
78
+ use, offer to sell, sell, import, and otherwise transfer the Work,
79
+ where such license applies only to those patent claims licensable
80
+ by such Contributor that are necessarily infringed by their
81
+ Contribution(s) alone or by combination of their Contribution(s)
82
+ with the Work to which such Contribution(s) was submitted. If You
83
+ institute patent litigation against any entity (including a
84
+ cross-claim or counterclaim in a lawsuit) alleging that the Work
85
+ or a Contribution incorporated within the Work constitutes direct
86
+ or contributory patent infringement, then any patent licenses
87
+ granted to You under this License for that Work shall terminate
88
+ as of the date such litigation is filed.
89
+
90
+ 4. Redistribution. You may reproduce and distribute copies of the
91
+ Work or Derivative Works thereof in any medium, with or without
92
+ modifications, and in Source or Object form, provided that You
93
+ meet the following conditions:
94
+
95
+ (a) You must give any other recipients of the Work or
96
+ Derivative Works a copy of this License; and
97
+
98
+ (b) You must cause any modified files to carry prominent notices
99
+ stating that You changed the files; and
100
+
101
+ (c) You must retain, in the Source form of any Derivative Works
102
+ that You distribute, all copyright, patent, trademark, and
103
+ attribution notices from the Source form of the Work,
104
+ excluding those notices that do not pertain to any part of
105
+ the Derivative Works; and
106
+
107
+ (d) If the Work includes a "NOTICE" text file as part of its
108
+ distribution, then any Derivative Works that You distribute must
109
+ include a readable copy of the attribution notices contained
110
+ within such NOTICE file, excluding those notices that do not
111
+ pertain to any part of the Derivative Works, in at least one
112
+ of the following places: within a NOTICE text file distributed
113
+ as part of the Derivative Works; within the Source form or
114
+ documentation, if provided along with the Derivative Works; or,
115
+ within a display generated by the Derivative Works, if and
116
+ wherever such third-party notices normally appear. The contents
117
+ of the NOTICE file are for informational purposes only and
118
+ do not modify the License. You may add Your own attribution
119
+ notices within Derivative Works that You distribute, alongside
120
+ or as an addendum to the NOTICE text from the Work, provided
121
+ that such additional attribution notices cannot be construed
122
+ as modifying the License.
123
+
124
+ You may add Your own copyright statement to Your modifications and
125
+ may provide additional or different license terms and conditions
126
+ for use, reproduction, or distribution of Your modifications, or
127
+ for any such Derivative Works as a whole, provided Your use,
128
+ reproduction, and distribution of the Work otherwise complies with
129
+ the conditions stated in this License.
130
+
131
+ 5. Submission of Contributions. Unless You explicitly state otherwise,
132
+ any Contribution intentionally submitted for inclusion in the Work
133
+ by You to the Licensor shall be under the terms and conditions of
134
+ this License, without any additional terms or conditions.
135
+ Notwithstanding the above, nothing herein shall supersede or modify
136
+ the terms of any separate license agreement you may have executed
137
+ with Licensor regarding such Contributions.
138
+
139
+ 6. Trademarks. This License does not grant permission to use the trade
140
+ names, trademarks, service marks, or product names of the Licensor,
141
+ except as required for reasonable and customary use in describing the
142
+ origin of the Work and reproducing the content of the NOTICE file.
143
+
144
+ 7. Disclaimer of Warranty. Unless required by applicable law or
145
+ agreed to in writing, Licensor provides the Work (and each
146
+ Contributor provides its Contributions) on an "AS IS" BASIS,
147
+ WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or
148
+ implied, including, without limitation, any warranties or conditions
149
+ of TITLE, NON-INFRINGEMENT, MERCHANTABILITY, or FITNESS FOR A
150
+ PARTICULAR PURPOSE. You are solely responsible for determining the
151
+ appropriateness of using or redistributing the Work and assume any
152
+ risks associated with Your exercise of permissions under this License.
153
+
154
+ 8. Limitation of Liability. In no event and under no legal theory,
155
+ whether in tort (including negligence), contract, or otherwise,
156
+ unless required by applicable law (such as deliberate and grossly
157
+ negligent acts) or agreed to in writing, shall any Contributor be
158
+ liable to You for damages, including any direct, indirect, special,
159
+ incidental, or consequential damages of any character arising as a
160
+ result of this License or out of the use or inability to use the
161
+ Work (including but not limited to damages for loss of goodwill,
162
+ work stoppage, computer failure or malfunction, or any and all
163
+ other commercial damages or losses), even if such Contributor
164
+ has been advised of the possibility of such damages.
165
+
166
+ 9. Accepting Warranty or Additional Liability. While redistributing
167
+ the Work or Derivative Works thereof, You may choose to offer,
168
+ and charge a fee for, acceptance of support, warranty, indemnity,
169
+ or other liability obligations and/or rights consistent with this
170
+ License. However, in accepting such obligations, You may act only
171
+ on Your own behalf and on Your sole responsibility, not on behalf
172
+ of any other Contributor, and only if You agree to indemnify,
173
+ defend, and hold each Contributor harmless for any liability
174
+ incurred by, or claims asserted against, such Contributor by reason
175
+ of your accepting any such warranty or additional liability.
176
+
177
+ END OF TERMS AND CONDITIONS
178
+
179
+ APPENDIX: How to apply the Apache License to your work.
180
+
181
+ To apply the Apache License to your work, attach the following
182
+ boilerplate notice, with the fields enclosed by brackets "[]"
183
+ replaced with your own identifying information. (Don't include
184
+ the brackets!) The text should be enclosed in the appropriate
185
+ comment syntax for the file format. We also recommend that a
186
+ file or class name and description of purpose be included on the
187
+ same "printed page" as the copyright notice for easier
188
+ identification within third-party archives.
189
+
190
+ Copyright [yyyy] [name of copyright owner]
191
+
192
+ Licensed under the Apache License, Version 2.0 (the "License");
193
+ you may not use this file except in compliance with the License.
194
+ You may obtain a copy of the License at
195
+
196
+ http://www.apache.org/licenses/LICENSE-2.0
197
+
198
+ Unless required by applicable law or agreed to in writing, software
199
+ distributed under the License is distributed on an "AS IS" BASIS,
200
+ WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
201
+ See the License for the specific language governing permissions and
202
+ limitations under the License.
@@ -0,0 +1,10 @@
1
+ Agent Comparability
2
+ Copyright 2026 Agent Comparability contributors
3
+
4
+ This product includes software developed as part of the Agent Comparability
5
+ project (specification cmp/0.3.1 and reference implementation).
6
+
7
+ This distribution bundles no third-party code. Optional integrations
8
+ (inspect-ai, pydantic-evals — both MIT) are user-installed and NOT bundled.
9
+ Conformance case data is original to this project.
10
+ This distribution contains NO CC BY-NC licensed data.
@@ -0,0 +1,351 @@
1
+ Metadata-Version: 2.4
2
+ Name: agent-comparability
3
+ Version: 0.5.2
4
+ Summary: Agent evaluation comparability semantics: spec + conformance suite + thin reference implementation
5
+ License: Apache-2.0
6
+ Requires-Python: >=3.10
7
+ Description-Content-Type: text/markdown
8
+ License-File: LICENSE
9
+ License-File: NOTICE
10
+ Provides-Extra: suite
11
+ Requires-Dist: pyyaml; extra == "suite"
12
+ Requires-Dist: jsonschema; extra == "suite"
13
+ Provides-Extra: inspect
14
+ Requires-Dist: inspect-ai; extra == "inspect"
15
+ Provides-Extra: pydantic
16
+ Requires-Dist: pydantic-evals; extra == "pydantic"
17
+ Dynamic: license-file
18
+
19
+ <div align="center">
20
+
21
+ # Agent Comparability
22
+
23
+ **Experimental Specification & Conformance Suite** for AI agent evaluation comparability
24
+
25
+ [![CI](https://github.com/Rethymus/agent-comparability/actions/workflows/ci.yml/badge.svg)](https://github.com/Rethymus/agent-comparability/actions/workflows/ci.yml)
26
+ [![License: Apache-2.0](https://img.shields.io/badge/License-Apache_2.0-blue.svg)](LICENSE)
27
+ [![Python 3.10+](https://img.shields.io/badge/python-3.10%2B-3776AB?logo=python&logoColor=white)](pyproject.toml)
28
+
29
+ **English** | [简体中文](README.zh-CN.md)
30
+
31
+ *When can you compare AI agent evaluation results — and what claims do they support?*
32
+
33
+ </div>
34
+
35
+ ---
36
+
37
+ > [!IMPORTANT]
38
+ > This project adjudicates **comparability of evaluation records** — it does **not** run
39
+ > agents, score benchmarks, or certify any agent's reliability. Passing the conformance
40
+ > suite means *specification* conformance, nothing more.
41
+
42
+ **The 30-second path**: [Install](#6-installation) → [Quickstart](#7-quickstart) →
43
+ [Real input/output](#8-inputoutput-example-actual-output).
44
+
45
+ <details open>
46
+ <summary><b>Table of Contents</b></summary>
47
+
48
+ 1. [Project Overview](#1-project-overview)
49
+ 2. [Problem & Motivation](#2-problem--motivation)
50
+ 3. [What Agent Comparability Does](#3-what-agent-comparability-does)
51
+ 4. [What It Does Not Do](#4-what-it-does-not-do)
52
+ 5. [Main Features](#5-main-features)
53
+ 6. [Installation](#6-installation)
54
+ 7. [Quickstart](#7-quickstart)
55
+ 8. [Input/Output Example (actual output)](#8-inputoutput-example-actual-output)
56
+ 9. [Architecture](#9-architecture)
57
+ 10. [Conformance Suite](#10-conformance-suite)
58
+ 11. [Inspect AI / Pydantic Evals Integrations](#11-inspect-ai--pydantic-evals-integrations)
59
+ 12. [Specification & Versioning](#12-specification--versioning)
60
+ 13. [Limitations](#13-limitations)
61
+ 14. [Contributing](#14-contributing)
62
+ 15. [Security](#15-security)
63
+ 16. [License & Acknowledgements](#16-license--acknowledgements)
64
+
65
+ [FAQ](#faq) · [Project status](#project-status)
66
+
67
+ </details>
68
+
69
+ ---
70
+
71
+ ## 1. Project Overview
72
+
73
+ Agent Comparability is a framework-independent **specification** ([spec/SPEC.md](spec/SPEC.md)),
74
+ a frozen **conformance suite**, and a thin **reference implementation** that adjudicates
75
+ *comparability* of agent evaluation records — never the scores themselves.
76
+
77
+ **Who is this for**: developers comparing agent/prompt/harness evaluation results,
78
+ benchmark authors documenting what conclusions their data supports, and teams building
79
+ reliability pipelines on top of existing eval frameworks.
80
+
81
+ ### How it relates to existing tools
82
+
83
+ | Tool | What it does | Relation to Agent Comparability |
84
+ |---|---|---|
85
+ | [Inspect AI](https://inspect.aisi.org.uk) | Runs evals, agents, sandboxes | Complementary — Inspect *executes*; this decides what comparisons its results support. Adapter provided. |
86
+ | [Pydantic Evals](https://pydantic.dev) | Datasets, cases, reporting | Complementary reporting layer over the same core. Adapter provided. |
87
+ | Benchmark harnesses (lm-eval, SWE-bench, …) | Measure models | They *produce* records; this adjudicates what comparisons of those records mean |
88
+
89
+ ## 2. Problem & Motivation
90
+
91
+ Three failure modes silently corrupt agent-evaluation conclusions:
92
+
93
+ | Failure mode | What actually goes wrong |
94
+ |---|---|
95
+ | **Semantics mixing** | A *verified failure* and an *unresolved zero* both read `0.0` — pooled "fail rates" become meaningless |
96
+ | **Confounded variables** | Model and harness change at the same time — a two-variable diff gets reported as causation |
97
+ | **Evidence gaps** | Missing traces quietly inflate success rates; rankings reverse across harnesses |
98
+
99
+ These boundaries should be **machine-checkable**, not vibes. That is what this project ships.
100
+
101
+ ## 3. What Agent Comparability Does
102
+
103
+ - Adjudicates a **Requested Claim** (e.g. `score_comparison`, `causal`,
104
+ `descriptive_registration`) against a set of evaluation records
105
+ - Returns a **Comparison Decision Record (CDR)**: admissibility
106
+ (`ADMISSIBLE / INADMISSIBLE / PARTIAL / INCONCLUSIVE`), satisfied/violated/unknown
107
+ conditions, **supported claims AND forbidden claims**, evidence refs, policy version
108
+ - Enforces evidence-trust tiers (agent-claimed vs source-verifier vs independently
109
+ verified), score-scale consistency (range/unit/verifier), zero-semantics splits,
110
+ and unknown-field/unknown-claim rules
111
+ - Ships a frozen conformance suite run against three separated metrics
112
+
113
+ **Admissibility quick reference** (full priority rules in [SPEC §3.3](spec/SPEC.md)):
114
+
115
+ | Verdict | Meaning | Typical trigger |
116
+ |---|---|---|
117
+ | `ADMISSIBLE` | Requested conclusion is supported | Same task / model / score scale, all records verified |
118
+ | `INADMISSIBLE` | Conclusion not supported | Confounded variables, task mismatch, or score-scale mismatch |
119
+ | `PARTIAL` | Only split presentation allowed | `verified` failures mixed with `unresolved_zero` |
120
+ | `INCONCLUSIVE` | Evidence insufficient to judge | Missing traces / evidence anchors |
121
+
122
+ ## 4. What It Does Not Do
123
+
124
+ > [!WARNING]
125
+ > - Does **not** run agents, benchmarks, or models; no network, no credentials, no Docker
126
+ > - Does **not** score anything or certify any agent's reliability
127
+ > - Never merges its three conformance metrics into one number (see [§10](#10-conformance-suite))
128
+
129
+ ## 5. Main Features
130
+
131
+ - `adjudicate(records, requested_claim) -> CDR` — pure Python, zero dependencies
132
+ - Priority-ordered decision table (evidence-missing > confounded > semantics > admissible)
133
+ - Machine-verifiable CDR JSON Schema ([spec/cdr.schema.json](spec/cdr.schema.json))
134
+ - Frozen A/B conformance case sets; B5 public erratum precedent protected
135
+ - Optional adapters for Inspect AI and Pydantic Evals reusing the same core
136
+
137
+ ## 6. Installation
138
+
139
+ From PyPI (first published release: `0.5.2`):
140
+
141
+ ```bash
142
+ pip install agent-comparability # minimal core, zero deps
143
+ pip install "agent-comparability[suite]" # + suite basics (pyyaml, jsonschema)
144
+ pip install "agent-comparability[suite,inspect,pydantic]" # FULL conformance run
145
+ ```
146
+
147
+ From a clone (development):
148
+
149
+ ```bash
150
+ git clone https://github.com/Rethymus/agent-comparability.git && cd agent-comparability
151
+ pip install -e . # minimal core, zero deps
152
+ pip install -e .[suite] # + suite basics (pyyaml, jsonschema)
153
+ pip install -e .[suite,inspect,pydantic] # FULL conformance run — required by run_suite.py
154
+ ```
155
+
156
+ Requires Python ≥ 3.10. No Docker, no model keys.
157
+
158
+ ## 7. Quickstart
159
+
160
+ ```python
161
+ from agent_comparability import adjudicate
162
+
163
+ cdr = adjudicate({"id": "demo", "records": [
164
+ {"task_id": "t1", "harness": "h1", "model": "m1",
165
+ "reward": 1.0, "reward_status": "verified"},
166
+ {"task_id": "t1", "harness": "h2", "model": "m1",
167
+ "reward": 0.0, "reward_status": "unresolved_zero"}]},
168
+ "score_comparison")
169
+ ```
170
+
171
+ ## 8. Input/Output Example (actual output)
172
+
173
+ > Note: the reference implementation's condition/claim messages are currently in Chinese; the admissibility verdict itself is locale-neutral.
174
+
175
+ The quickstart call returns:
176
+
177
+ ```
178
+ cdr.admissibility -> PARTIAL
179
+ cdr.supported_claims -> ['per_status_stratified_rate(verified_zero/unresolved_zero 分列)']
180
+ ```
181
+
182
+ A confounded group instead:
183
+
184
+ ```python
185
+ cdr = adjudicate({"id": "x", "records": [
186
+ {"task_id": "a", "harness": "h1", "model": "m1", "reward": 1.0, "reward_status": "verified"},
187
+ {"task_id": "b", "harness": "h2", "model": "m2", "reward": 0.0, "reward_status": "verified"}]},
188
+ "score_comparison")
189
+ ```
190
+
191
+ ```
192
+ cdr.admissibility -> INADMISSIBLE
193
+ cdr.violated_conditions -> ['same_task_id', 'same_model', 'single_variable_isolation']
194
+ ```
195
+
196
+ `cdr.to_dict()` emits the full CDR (schema-validated). The quickstart call's complete
197
+ CDR — actual output:
198
+
199
+ ```json
200
+ {
201
+ "case_id": "demo",
202
+ "requested_claim": "score_comparison",
203
+ "comparison_scope": "2 records; models=['m1']; harnesses=['h1', 'h2']",
204
+ "admissibility": "PARTIAL",
205
+ "satisfied_conditions": [
206
+ "same_task_id",
207
+ "same_model",
208
+ "single_variable_isolation",
209
+ "continuous_score_preserved({'range': '0..1', 'unit': 'fraction', 'verifier': 'source'})"
210
+ ],
211
+ "violated_conditions": [
212
+ "zero_semantics_uniform(verified_zero != unresolved_zero)"
213
+ ],
214
+ "unknown_conditions": [],
215
+ "supported_claims": [
216
+ "per_status_stratified_rate(verified_zero/unresolved_zero 分列)"
217
+ ],
218
+ "unsupported_claims": [
219
+ "pooled_fail_rate(0 分语义混同)"
220
+ ],
221
+ "evidence_refs": [
222
+ "rec:t1/m1/h1",
223
+ "rec:t1/m1/h2"
224
+ ],
225
+ "policy_version": "cmp/0.3.1",
226
+ "limitations": [
227
+ "摘要一致仅证明未同步修改,不构成抗恶意篡改(单写信任域)"
228
+ ]
229
+ }
230
+ ```
231
+
232
+ ## 9. Architecture
233
+
234
+ ```
235
+ spec/SPEC.md ← single source of truth (readable standalone)
236
+ spec/cdr.schema.json ← machine-verifiable CDR
237
+ src/agent_comparability/core.py ← thin reference implementation
238
+ src/agent_comparability/data/ ← SPEC + schema + frozen cases (packaged)
239
+ conformance/ ← run_suite.py (3 metrics) + independent_impl.py + cases
240
+ adapters/ ← optional inspect-ai / pydantic-evals wrappers
241
+ tests/ ← smoke + property tests
242
+ ```
243
+
244
+ Adjudication flow (priority order, per [SPEC §3.3](spec/SPEC.md)):
245
+
246
+ ```mermaid
247
+ flowchart LR
248
+ A[Records + Requested Claim] --> B{trace available?}
249
+ B -- missing --> C[INCONCLUSIVE]
250
+ B -- ok --> D{task / model /<br/>score scale consistent?}
251
+ D -- conflict --> E[INADMISSIBLE]
252
+ D -- ok --> F{zero semantics uniform?}
253
+ F -- verified + unresolved mixed --> G[PARTIAL]
254
+ F -- ok --> H[ADMISSIBLE]
255
+ C & E & G & H --> V[CDR<br/>supported + forbidden claims]
256
+ ```
257
+
258
+ ## 10. Conformance Suite
259
+
260
+ `python conformance/run_suite.py` reports three **separated** metrics that must
261
+ never be merged:
262
+
263
+ 1. **Policy correctness** — reference implementation vs frozen expectations
264
+ 2. **Spec conformance** — an independent implementation written from SPEC alone
265
+ 3. **Integration consistency** — the same core via Inspect/Pydantic adapters
266
+
267
+ A failure in each means a different defect class (implementation bug / spec
268
+ ambiguity or misread / adapter bug).
269
+
270
+ The full suite run requires the adapter extras (`.[suite,inspect,pydantic]`) — a missing adapter is treated as FAIL, not skipped.
271
+
272
+ ## 11. Inspect AI / Pydantic Evals Integrations
273
+
274
+ ```bash
275
+ pip install -e .[inspect] && python adapters/inspect_adapter.py
276
+ pip install -e .[pydantic] && python adapters/pydantic_adapter.py
277
+ ```
278
+
279
+ Both wrap the *same* core (no semantic duplication). Inspect scorer uses
280
+ `metrics=[]` — admissibility is deliberately **not** aggregated into accuracy.
281
+ The Pydantic evaluator checks the **task's real output** (`ctx.output`) against
282
+ frozen expectations — it never re-adjudicates inside the evaluator — and the
283
+ adapter's exit code derives from the full 22-pair `report` (cases + failures).
284
+
285
+ ## 12. Specification & Versioning
286
+
287
+ Three distinct lines: software tag (e.g. `v0.5.0`), specification `cmp/0.3.1`
288
+ (inside SPEC), and `agent_comparability.__version__`. Judgment-table changes
289
+ require a spec major bump (SPEC §5). See [CHANGELOG](CHANGELOG.md) for the full tag lineage
290
+ (v0.3.1–v0.3.6 are deprecated candidates kept immutably for the record).
291
+
292
+ ## 13. Limitations
293
+
294
+ - Experimental specification; self-tested conformance, **no external certification**
295
+ - Single-writer trust domain: digest equality proves non-divergence, not
296
+ tamper-resistance (SPEC §1)
297
+ - No repetition/cost modeling inside adjudication (handled by callers)
298
+ - Chinese/English docs kept equivalent; translation drift tracked in CI link checks
299
+
300
+ ## FAQ
301
+
302
+ **Does passing the conformance suite certify my agent is reliable?**
303
+ No. It certifies that an *implementation* matches the specification. It says nothing about any agent.
304
+
305
+ **Can I use it with my existing eval framework?**
306
+ Yes. The core is framework-independent (`adjudicate()` is a pure function over record
307
+ dicts). Adapters for Inspect AI and Pydantic Evals are included; any framework whose
308
+ output you can map to the [record schema](spec/SPEC.md#0-范围与目的) works.
309
+
310
+ **Why not just compare scores?**
311
+ Because `0.0` can mean a *verified failure* or an *unresolved zero*; because model and
312
+ harness changing together turns a diff into an unfounded causal claim; and because
313
+ missing traces inflate success rates silently. §2's failure modes are all real-world
314
+ artifacts, not hypotheticals.
315
+
316
+ **Does it run agents or benchmarks?**
317
+ No — no network, no credentials, no Docker. It only ever sees the records you pass in.
318
+
319
+ **Which versions should I depend on?**
320
+ Package `v0.5.0+` (version lines unified and CI-enforced since then); specification
321
+ `cmp/0.3.1` evolves independently — see [§12](#12-specification--versioning).
322
+
323
+ ## Project status
324
+
325
+ | Component | Status |
326
+ |---|---|
327
+ | Specification `cmp/0.3.1` | Stable — frozen conformance set anchors it |
328
+ | Claim types | `score_comparison`, `causal`, `descriptive_registration` (+ unknown → `INCONCLUSIVE`) |
329
+ | Conformance cases | 22 frozen pairs (A + B sets), three separated metrics, all green in CI |
330
+ | Adapters | Inspect AI ✓ · Pydantic Evals ✓ |
331
+ | CI gates | Terminal scan · bilingual docs · conformance ×3 · tests · release version · build + twine · package content |
332
+ | Repo governance | `main` protected (PR + required check), release tags `v*` immutable |
333
+ | Distribution | PyPI (`agent-comparability`, since `0.5.2`, Trusted Publishing) + GitHub source/tags |
334
+
335
+ ## 14. Contributing
336
+
337
+ See [CONTRIBUTING.md](CONTRIBUTING.md). Key rule: **never edit frozen expectations
338
+ to make tests pass**; errata follow the public B5 precedent. PRs must pass all
339
+ three conformance metrics separately.
340
+
341
+ ## 15. Security
342
+
343
+ See [SECURITY.md](SECURITY.md). Private vulnerability reporting preferred.
344
+
345
+ ## 16. License & Acknowledgements
346
+
347
+ Apache-2.0 (see [LICENSE](LICENSE), [NOTICE](NOTICE)). Optional integrations
348
+ inspect-ai / pydantic-evals are MIT and **not bundled**. Conformance case data is
349
+ original to this project; no CC BY-NC data is distributed. Research lineage:
350
+ this repo was distilled from a private research repository; internal process
351
+ records live under `docs/internal/`.