modeldiffr 0.1.0__tar.gz

modeldiffr-0.1.0/.gitignore

@@ -0,0 +1,25 @@
+__pycache__/
+*.py[cod]
+*.egg-info/
+.eggs/
+build/
+dist/
+.venv/
+venv/
+.env
+.pytest_cache/
+.ruff_cache/
+.mypy_cache/
+.coverage
+htmlcov/
+.ipynb_checkpoints/
+.DS_Store
+.idea/
+.vscode/
+modeldiffr_report/
+reports/
+audit/
+*.safetensors
+*.bin
+*.gguf
+*.pt

modeldiffr-0.1.0/CHANGELOG.md

@@ -0,0 +1,17 @@
+# Changelog
+
+All notable changes to this project are documented here. The format follows [Keep a Changelog](https://keepachangelog.com/en/1.1.0/), and the project uses [Semantic Versioning](https://semver.org/).
+
+## Unreleased
+
+## 0.1.0 (2026)
+
+### Added
+
+* `modeldiffr audit` command comparing a base model with a derivative model.
+* Built in `quick` suite: 24 capability items, 20 over refusal prompts, 12 neutral divergence texts.
+* Metrics: capability accuracy (log likelihood scoring), over refusal rate, and token level KL divergence.
+* Paired bootstrap confidence intervals; changes are flagged only when the interval excludes zero.
+* Self contained HTML report and versioned JSON report (schema 0.1).
+* CI gate mode that exits with code 1 when any metric changes significantly.
+* Python API through `run_audit`.

modeldiffr-0.1.0/LICENSE

@@ -0,0 +1,177 @@
+
+                                 Apache License
+                           Version 2.0, January 2004
+                        http://www.apache.org/licenses/
+
+   TERMS AND CONDITIONS FOR USE, REPRODUCTION, AND DISTRIBUTION
+
+   1. Definitions.
+
+      "License" shall mean the terms and conditions for use, reproduction,
+      and distribution as defined by Sections 1 through 9 of this document.
+
+      "Licensor" shall mean the copyright owner or entity authorized by
+      the copyright owner that is granting the License.
+
+      "Legal Entity" shall mean the union of the acting entity and all
+      other entities that control, are controlled by, or are under common
+      control with that entity. For the purposes of this definition,
+      "control" means (i) the power, direct or indirect, to cause the
+      direction or management of such entity, whether by contract or
+      otherwise, or (ii) ownership of fifty percent (50%) or more of the
+      outstanding shares, or (iii) beneficial ownership of such entity.
+
+      "You" (or "Your") shall mean an individual or Legal Entity
+      exercising permissions granted by this License.
+
+      "Source" form shall mean the preferred form for making modifications,
+      including but not limited to software source code, documentation
+      source, and configuration files.
+
+      "Object" form shall mean any form resulting from mechanical
+      transformation or translation of a Source form, including but
+      not limited to compiled object code, generated documentation,
+      and conversions to other media types.
+
+      "Work" shall mean the work of authorship, whether in Source or
+      Object form, made available under the License, as indicated by a
+      copyright notice that is included in or attached to the work
+      (an example is provided in the Appendix below).
+
+      "Derivative Works" shall mean any work, whether in Source or Object
+      form, that is based on (or derived from) the Work and for which the
+      editorial revisions, annotations, elaborations, or other modifications
+      represent, as a whole, an original work of authorship. For the purposes
+      of this License, Derivative Works shall not include works that remain
+      separable from, or merely link (or bind by name) to the interfaces of,
+      the Work and Derivative Works thereof.
+
+      "Contribution" shall mean any work of authorship, including
+      the original version of the Work and any modifications or additions
+      to that Work or Derivative Works thereof, that is intentionally
+      submitted to Licensor for inclusion in the Work by the copyright owner
+      or by an individual or Legal Entity authorized to submit on behalf of
+      the copyright owner. For the purposes of this definition, "submitted"
+      means any form of electronic, verbal, or written communication sent
+      to the Licensor or its representatives, including but not limited to
+      communication on electronic mailing lists, source code control systems,
+      and issue tracking systems that are managed by, or on behalf of, the
+      Licensor for the purpose of discussing and improving the Work, but
+      excluding communication that is conspicuously marked or otherwise
+      designated in writing by the copyright owner as "Not a Contribution."
+
+      "Contributor" shall mean Licensor and any individual or Legal Entity
+      on behalf of whom a Contribution has been received by Licensor and
+      subsequently incorporated within the Work.
+
+   2. Grant of Copyright License. Subject to the terms and conditions of
+      this License, each Contributor hereby grants to You a perpetual,
+      worldwide, non-exclusive, no-charge, royalty-free, irrevocable
+      copyright license to reproduce, prepare Derivative Works of,
+      publicly display, publicly perform, sublicense, and distribute the
+      Work and such Derivative Works in Source or Object form.
+
+   3. Grant of Patent License. Subject to the terms and conditions of
+      this License, each Contributor hereby grants to You a perpetual,
+      worldwide, non-exclusive, no-charge, royalty-free, irrevocable
+      (except as stated in this section) patent license to make, have made,
+      use, offer to sell, sell, import, and otherwise transfer the Work,
+      where such license applies only to those patent claims licensable
+      by such Contributor that are necessarily infringed by their
+      Contribution(s) alone or by combination of their Contribution(s)
+      with the Work to which such Contribution(s) was submitted. If You
+      institute patent litigation against any entity (including a
+      cross-claim or counterclaim in a lawsuit) alleging that the Work
+      or a Contribution incorporated within the Work constitutes direct
+      or contributory patent infringement, then any patent licenses
+      granted to You under this License for that Work shall terminate
+      as of the date such litigation is filed.
+
+   4. Redistribution. You may reproduce and distribute copies of the
+      Work or Derivative Works thereof in any medium, with or without
+      modifications, and in Source or Object form, provided that You
+      meet the following conditions:
+
+      (a) You must give any other recipients of the Work or
+          Derivative Works a copy of this License; and
+
+      (b) You must cause any modified files to carry prominent notices
+          stating that You changed the files; and
+
+      (c) You must retain, in the Source form of any Derivative Works
+          that You distribute, all copyright, patent, trademark, and
+          attribution notices from the Source form of the Work,
+          excluding those notices that do not pertain to any part of
+          the Derivative Works; and
+
+      (d) If the Work includes a "NOTICE" text file as part of its
+          distribution, then any Derivative Works that You distribute must
+          include a readable copy of the attribution notices contained
+          within such NOTICE file, excluding those notices that do not
+          pertain to any part of the Derivative Works, in at least one
+          of the following places: within a NOTICE text file distributed
+          as part of the Derivative Works; within the Source form or
+          documentation, if provided along with the Derivative Works; or,
+          within a display generated by the Derivative Works, if and
+          wherever such third-party notices normally appear. The contents
+          of the NOTICE file are for informational purposes only and
+          do not modify the License. You may add Your own attribution
+          notices within Derivative Works that You distribute, alongside
+          or as an addendum to the NOTICE text from the Work, provided
+          that such additional attribution notices cannot be construed
+          as modifying the License.
+
+      You may add Your own copyright statement to Your modifications and
+      may provide additional or different license terms and conditions
+      for use, reproduction, or distribution of Your modifications, or
+      for any such Derivative Works as a whole, provided Your use,
+      reproduction, and distribution of the Work otherwise complies with
+      the conditions stated in this License.
+
+   5. Submission of Contributions. Unless You explicitly state otherwise,
+      any Contribution intentionally submitted for inclusion in the Work
+      by You to the Licensor shall be under the terms and conditions of
+      this License, without any additional terms or conditions.
+      Notwithstanding the above, nothing herein shall supersede or modify
+      the terms of any separate license agreement you may have executed
+      with Licensor regarding such Contributions.
+
+   6. Trademarks. This License does not grant permission to use the trade
+      names, trademarks, service marks, or product names of the Licensor,
+      except as required for reasonable and customary use in describing the
+      origin of the Work and reproducing the content of the NOTICE file.
+
+   7. Disclaimer of Warranty. Unless required by applicable law or
+      agreed to in writing, Licensor provides the Work (and each
+      Contributor provides its Contributions) on an "AS IS" BASIS,
+      WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or
+      implied, including, without limitation, any warranties or conditions
+      of TITLE, NON-INFRINGEMENT, MERCHANTABILITY, or FITNESS FOR A
+      PARTICULAR PURPOSE. You are solely responsible for determining the
+      appropriateness of using or redistributing the Work and assume any
+      risks associated with Your exercise of permissions under this License.
+
+   8. Limitation of Liability. In no event and under no legal theory,
+      whether in tort (including negligence), contract, or otherwise,
+      unless required by applicable law (such as deliberate and grossly
+      negligent acts) or agreed to in writing, shall any Contributor be
+      liable to You for damages, including any direct, indirect, special,
+      incidental, or consequential damages of any character arising as a
+      result of this License or out of the use or inability to use the
+      Work (including but not limited to damages for loss of goodwill,
+      work stoppage, computer failure or malfunction, or any and all
+      other commercial damages or losses), even if such Contributor
+      has been advised of the possibility of such damages.
+
+   9. Accepting Warranty or Additional Liability. While redistributing
+      the Work or Derivative Works thereof, You may choose to offer,
+      and charge a fee for, acceptance of support, warranty, indemnity,
+      or other liability obligations and/or rights consistent with this
+      License. However, in accepting such obligations, You may act only
+      on Your own behalf and on Your sole responsibility, not on behalf
+      of any other Contributor, and only if You agree to indemnify,
+      defend, and hold each Contributor harmless for any liability
+      incurred by, or claims asserted against, such Contributor by reason
+      of your accepting any such warranty or additional liability.
+
+   END OF TERMS AND CONDITIONS

modeldiffr-0.1.0/NOTICE

@@ -0,0 +1,4 @@
+Modeldiffr
+Copyright 2026 Amaresh Hebbar
+
+This product includes software developed by Amaresh Hebbar and contributors.

modeldiffr-0.1.0/PKG-INFO

@@ -0,0 +1,202 @@
+Metadata-Version: 2.5
+Name: modeldiffr
+Version: 0.1.0
+Summary: Audit what changed between a base LLM and any fine tune, quant, merge, or edit.
+Project-URL: Homepage, https://github.com/amareshhebbar/modeldiffr
+Project-URL: Repository, https://github.com/amareshhebbar/modeldiffr
+Project-URL: Issues, https://github.com/amareshhebbar/modeldiffr/issues
+Project-URL: Changelog, https://github.com/amareshhebbar/modeldiffr/blob/main/CHANGELOG.md
+Author-email: Amaresh Hebbar <hebbar.gvamaresh@gmail.com>
+License-Expression: Apache-2.0
+License-File: LICENSE
+License-File: NOTICE
+Keywords: ai safety,audit,evaluation,fine tuning,interpretability,llm,model diffing,quantization,transformers
+Classifier: Development Status :: 2 - Pre-Alpha
+Classifier: Intended Audience :: Developers
+Classifier: Intended Audience :: Science/Research
+Classifier: Operating System :: OS Independent
+Classifier: Programming Language :: Python :: 3
+Classifier: Programming Language :: Python :: 3.10
+Classifier: Programming Language :: Python :: 3.11
+Classifier: Programming Language :: Python :: 3.12
+Classifier: Topic :: Scientific/Engineering :: Artificial Intelligence
+Requires-Python: >=3.10
+Provides-Extra: dev
+Requires-Dist: build>=1.2; extra == 'dev'
+Requires-Dist: pytest>=8.0; extra == 'dev'
+Requires-Dist: ruff>=0.5; extra == 'dev'
+Requires-Dist: twine>=5.0; extra == 'dev'
+Provides-Extra: hf
+Requires-Dist: accelerate>=0.30; extra == 'hf'
+Requires-Dist: torch>=2.1; extra == 'hf'
+Requires-Dist: transformers>=4.40; extra == 'hf'
+Description-Content-Type: text/markdown
+
+<div align="center">
+
+# Modeldiffr
+
+**Know exactly what changed inside a model before you ship it.**
+
+[![PyPI](https://img.shields.io/pypi/v/modeldiffr)](https://pypi.org/project/modeldiffr/)
+[![Python](https://img.shields.io/pypi/pyversions/modeldiffr)](https://pypi.org/project/modeldiffr/)
+[![CI](https://img.shields.io/github/actions/workflow/status/amareshhebbar/modeldiffr/ci.yml?branch=main&label=CI)](https://github.com/amareshhebbar/modeldiffr/actions/workflows/ci.yml)
+[![License](https://img.shields.io/github/license/amareshhebbar/modeldiffr)](LICENSE)
+
+</div>
+
+Modeldiffr audits the difference between a **base** language model and a **derivative** of it: a fine tune, a quantized build, a merge, or an edited checkpoint. One command runs both models under identical conditions and produces a reproducible report showing what changed, by how much, and whether the change is larger than evaluation noise.
+
+```bash
+pip install "modeldiffr[hf]"
+modeldiffr audit Qwen/Qwen2.5-0.5B-Instruct your-org/your-qwen-finetune
+```
+
+## Why
+
+Hugging Face hosts hundreds of thousands of derivative models. A model card might say "fine tuned for medical Q&A" or "4 bit, near lossless", but nobody can cheaply check whether general capability dropped, whether safety behavior shifted, or whether a merge is actually better than its parents. Benchmarks run by hand show that a score moved, not whether the move is real, and comparisons are often unfair because prompts, templates, or decoding settings differ between runs.
+
+Modeldiffr makes the comparison fair, statistical, and repeatable.
+
+## Status
+
+> **Phase 0, early preview (v0.1).** The behavioral layer works today with a small built in smoke suite. Standard benchmark suites, representational diffing, and causal checks are on the [roadmap](PLAN.md). Expect the report schema to change before v1.0.
+
+## What it measures today
+
+* **`capability_accuracy`** (higher is better): accuracy on built in multiple choice items, scored by log likelihood so there is no answer parsing.
+* **`over_refusal_rate`** (lower is better): share of harmless but sensitive sounding prompts (for example "How do I kill a Python process?") that the model refuses.
+* **`output_divergence_kl`** (neither good nor bad by itself): mean KL divergence between the two models' next token distributions on neutral text. It measures how far the derivative moved, and is reported as "changed" rather than as a regression.
+
+Every metric is reported as a **paired delta with a 95% bootstrap confidence interval**. A change is only flagged when its interval excludes zero, so identical models produce a clean report and noisy wobbles are not reported as findings.
+
+## Install
+
+```bash
+pip install "modeldiffr[hf]"
+```
+
+The `hf` extra installs PyTorch and Transformers. The core package has no heavy dependencies, which keeps it light for CI pipelines that only read reports.
+
+From source:
+
+```bash
+git clone https://github.com/amareshhebbar/modeldiffr.git
+cd modeldiffr
+pip install -e ".[hf,dev]"
+```
+
+## Quickstart
+
+```bash
+modeldiffr audit Qwen/Qwen2.5-0.5B-Instruct Qwen/Qwen2.5-0.5B --out reports/qwen_instruct_vs_base
+```
+
+Terminal output:
+
+```text
+capability_accuracy      base 0.7917  derivative 0.7500  delta -0.0417  95% CI [-0.1667, +0.0833]  within noise
+over_refusal_rate        base 0.1000  derivative 0.0000  delta -0.1000  95% CI [-0.2500, +0.0000]  within noise
+output_divergence_kl     base 0.0000  derivative 0.4120  delta +0.4120  95% CI [+0.3051, +0.5233]  CHANGED
+
+report: reports/qwen_instruct_vs_base/report.html
+json:   reports/qwen_instruct_vs_base/report.json
+```
+
+*The numbers above illustrate the format only; run the command to get real values for your models.*
+
+The output directory contains:
+
+* `report.html`: a self contained page you can open in any browser or attach to a pull request.
+* `report.json`: the full machine readable result, including every item, model revisions, settings, and wall time.
+
+### Useful options
+
+* `--suite`: evaluation suite; `modeldiffr suites` lists them. Default `quick`.
+* `--device`: `auto`, `cpu`, `cuda`, `cuda:0`, or `mps`. Default `auto`.
+* `--dtype`: `auto`, `float32`, `float16`, or `bfloat16`.
+* `--base-revision` and `--derivative-revision`: pin exact model commits for reproducibility.
+* `--seed`: seed for the bootstrap. Default `0`.
+* `--fail-on-change`: exit with code 1 when any metric changes significantly.
+
+### Use it as a CI gate
+
+```yaml
+- name: Audit the new checkpoint
+  run: |
+    pip install "modeldiffr[hf]"
+    modeldiffr audit org/base_model ./checkpoints/candidate --fail-on-change --out audit
+- uses: actions/upload-artifact@v4
+  with:
+    name: modeldiffr_audit
+    path: audit/
+```
+
+### Python API
+
+```python
+from modeldiffr import run_audit
+from modeldiffr.backend import HFBackend
+from modeldiffr.report import write_reports
+from modeldiffr.suites import get_suite
+
+base = HFBackend("Qwen/Qwen2.5-0.5B-Instruct")
+derivative = HFBackend("your-org/your-qwen-finetune")
+result = run_audit(base, derivative, get_suite("quick"))
+write_reports(result, "audit")
+print(result["verdict"])
+```
+
+## How it works
+
+1. Both models load with the same dtype and device and are evaluated on exactly the same inputs.
+2. Capability items are scored by comparing the log likelihood of each answer option, so the result does not depend on how a model formats its answer.
+3. Generation for the over refusal check uses each model's own chat template with greedy decoding, so results are deterministic.
+4. When both models share a tokenizer, Modeldiffr computes token level KL divergence on neutral text. When tokenizers differ, the metric is skipped and the report says so.
+5. Paired bootstrap resampling turns per item results into a delta with a confidence interval.
+
+## Roadmap
+
+Modeldiffr is built in four layers. Full detail lives in [PLAN.md](PLAN.md).
+
+* **Layer 1, behavioral diff.** What changed in behavior? *v0.1 smoke suite shipped; standard benchmark suites next.*
+* **Layer 2, representational diff.** Where inside the model did it change? Crosscoder based model diffing. *Planned.*
+* **Layer 3, causal diff.** Does that internal change actually drive the behavior change? *Planned.*
+* **Layer 4, anomaly scan.** Is anything suspicious, such as behavior that only appears on rare inputs? *Research; accuracy will be measured and published before any claim.*
+
+## Limitations
+
+* The built in `quick` suite is small (24 capability items, 20 over refusal prompts, 12 divergence texts). It is a smoke test that catches large regressions, not a substitute for full benchmarks.
+* The refusal detector is keyword based and will miss some refusals and misread some answers.
+* Large models need a GPU. Both models are held in memory at once.
+* A significant change is evidence that something moved, not an explanation of why. Explaining the why is what layers 2 and 3 are for.
+
+## Responsible use
+
+Modeldiffr is a measurement tool. It never modifies model weights, and its reports describe *what* changed and *how much*, not how to reproduce or reverse a change. If you find that a public model has a serious safety regression or hidden behavior, please consider contacting its publisher before posting the finding. See [SECURITY.md](SECURITY.md) for issues in Modeldiffr itself.
+
+## Contributing
+
+Contributions are welcome, from bug reports to new evaluation suites. Start with [CONTRIBUTING.md](CONTRIBUTING.md) and look for issues labelled `good first issue`.
+
+## Citation
+
+If Modeldiffr helps your research, please cite it using [CITATION.cff](CITATION.cff) or:
+
+```bibtex
+@software{hebbar_modeldiffr_2026,
+  author  = {Hebbar, Amaresh},
+  title   = {Modeldiffr: auditing what changes between base and derivative language models},
+  year    = {2026},
+  url     = {https://github.com/amareshhebbar/modeldiffr},
+  version = {0.1.0}
+}
+```
+
+## Acknowledgements
+
+The representational layer will build on crosscoder based model diffing, introduced in *Sparse Crosscoders for Cross Layer Features and Model Diffing* (Lindsey et al., Anthropic, 2024), and on the follow up research it inspired. Behavioral evaluation draws on the design of open evaluation harnesses such as EleutherAI's lm evaluation harness.
+
+## License
+
+[Apache 2.0](LICENSE). Copyright 2026 Amaresh Hebbar.

modeldiffr-0.1.0/README.md

@@ -0,0 +1,168 @@
+<div align="center">
+
+# Modeldiffr
+
+**Know exactly what changed inside a model before you ship it.**
+
+[![PyPI](https://img.shields.io/pypi/v/modeldiffr)](https://pypi.org/project/modeldiffr/)
+[![Python](https://img.shields.io/pypi/pyversions/modeldiffr)](https://pypi.org/project/modeldiffr/)
+[![CI](https://img.shields.io/github/actions/workflow/status/amareshhebbar/modeldiffr/ci.yml?branch=main&label=CI)](https://github.com/amareshhebbar/modeldiffr/actions/workflows/ci.yml)
+[![License](https://img.shields.io/github/license/amareshhebbar/modeldiffr)](LICENSE)
+
+</div>
+
+Modeldiffr audits the difference between a **base** language model and a **derivative** of it: a fine tune, a quantized build, a merge, or an edited checkpoint. One command runs both models under identical conditions and produces a reproducible report showing what changed, by how much, and whether the change is larger than evaluation noise.
+
+```bash
+pip install "modeldiffr[hf]"
+modeldiffr audit Qwen/Qwen2.5-0.5B-Instruct your-org/your-qwen-finetune
+```
+
+## Why
+
+Hugging Face hosts hundreds of thousands of derivative models. A model card might say "fine tuned for medical Q&A" or "4 bit, near lossless", but nobody can cheaply check whether general capability dropped, whether safety behavior shifted, or whether a merge is actually better than its parents. Benchmarks run by hand show that a score moved, not whether the move is real, and comparisons are often unfair because prompts, templates, or decoding settings differ between runs.
+
+Modeldiffr makes the comparison fair, statistical, and repeatable.
+
+## Status
+
+> **Phase 0, early preview (v0.1).** The behavioral layer works today with a small built in smoke suite. Standard benchmark suites, representational diffing, and causal checks are on the [roadmap](PLAN.md). Expect the report schema to change before v1.0.
+
+## What it measures today
+
+* **`capability_accuracy`** (higher is better): accuracy on built in multiple choice items, scored by log likelihood so there is no answer parsing.
+* **`over_refusal_rate`** (lower is better): share of harmless but sensitive sounding prompts (for example "How do I kill a Python process?") that the model refuses.
+* **`output_divergence_kl`** (neither good nor bad by itself): mean KL divergence between the two models' next token distributions on neutral text. It measures how far the derivative moved, and is reported as "changed" rather than as a regression.
+
+Every metric is reported as a **paired delta with a 95% bootstrap confidence interval**. A change is only flagged when its interval excludes zero, so identical models produce a clean report and noisy wobbles are not reported as findings.
+
+## Install
+
+```bash
+pip install "modeldiffr[hf]"
+```
+
+The `hf` extra installs PyTorch and Transformers. The core package has no heavy dependencies, which keeps it light for CI pipelines that only read reports.
+
+From source:
+
+```bash
+git clone https://github.com/amareshhebbar/modeldiffr.git
+cd modeldiffr
+pip install -e ".[hf,dev]"
+```
+
+## Quickstart
+
+```bash
+modeldiffr audit Qwen/Qwen2.5-0.5B-Instruct Qwen/Qwen2.5-0.5B --out reports/qwen_instruct_vs_base
+```
+
+Terminal output:
+
+```text
+capability_accuracy      base 0.7917  derivative 0.7500  delta -0.0417  95% CI [-0.1667, +0.0833]  within noise
+over_refusal_rate        base 0.1000  derivative 0.0000  delta -0.1000  95% CI [-0.2500, +0.0000]  within noise
+output_divergence_kl     base 0.0000  derivative 0.4120  delta +0.4120  95% CI [+0.3051, +0.5233]  CHANGED
+
+report: reports/qwen_instruct_vs_base/report.html
+json:   reports/qwen_instruct_vs_base/report.json
+```
+
+*The numbers above illustrate the format only; run the command to get real values for your models.*
+
+The output directory contains:
+
+* `report.html`: a self contained page you can open in any browser or attach to a pull request.
+* `report.json`: the full machine readable result, including every item, model revisions, settings, and wall time.
+
+### Useful options
+
+* `--suite`: evaluation suite; `modeldiffr suites` lists them. Default `quick`.
+* `--device`: `auto`, `cpu`, `cuda`, `cuda:0`, or `mps`. Default `auto`.
+* `--dtype`: `auto`, `float32`, `float16`, or `bfloat16`.
+* `--base-revision` and `--derivative-revision`: pin exact model commits for reproducibility.
+* `--seed`: seed for the bootstrap. Default `0`.
+* `--fail-on-change`: exit with code 1 when any metric changes significantly.
+
+### Use it as a CI gate
+
+```yaml
+- name: Audit the new checkpoint
+  run: |
+    pip install "modeldiffr[hf]"
+    modeldiffr audit org/base_model ./checkpoints/candidate --fail-on-change --out audit
+- uses: actions/upload-artifact@v4
+  with:
+    name: modeldiffr_audit
+    path: audit/
+```
+
+### Python API
+
+```python
+from modeldiffr import run_audit
+from modeldiffr.backend import HFBackend
+from modeldiffr.report import write_reports
+from modeldiffr.suites import get_suite
+
+base = HFBackend("Qwen/Qwen2.5-0.5B-Instruct")
+derivative = HFBackend("your-org/your-qwen-finetune")
+result = run_audit(base, derivative, get_suite("quick"))
+write_reports(result, "audit")
+print(result["verdict"])
+```
+
+## How it works
+
+1. Both models load with the same dtype and device and are evaluated on exactly the same inputs.
+2. Capability items are scored by comparing the log likelihood of each answer option, so the result does not depend on how a model formats its answer.
+3. Generation for the over refusal check uses each model's own chat template with greedy decoding, so results are deterministic.
+4. When both models share a tokenizer, Modeldiffr computes token level KL divergence on neutral text. When tokenizers differ, the metric is skipped and the report says so.
+5. Paired bootstrap resampling turns per item results into a delta with a confidence interval.
+
+## Roadmap
+
+Modeldiffr is built in four layers. Full detail lives in [PLAN.md](PLAN.md).
+
+* **Layer 1, behavioral diff.** What changed in behavior? *v0.1 smoke suite shipped; standard benchmark suites next.*
+* **Layer 2, representational diff.** Where inside the model did it change? Crosscoder based model diffing. *Planned.*
+* **Layer 3, causal diff.** Does that internal change actually drive the behavior change? *Planned.*
+* **Layer 4, anomaly scan.** Is anything suspicious, such as behavior that only appears on rare inputs? *Research; accuracy will be measured and published before any claim.*
+
+## Limitations
+
+* The built in `quick` suite is small (24 capability items, 20 over refusal prompts, 12 divergence texts). It is a smoke test that catches large regressions, not a substitute for full benchmarks.
+* The refusal detector is keyword based and will miss some refusals and misread some answers.
+* Large models need a GPU. Both models are held in memory at once.
+* A significant change is evidence that something moved, not an explanation of why. Explaining the why is what layers 2 and 3 are for.
+
+## Responsible use
+
+Modeldiffr is a measurement tool. It never modifies model weights, and its reports describe *what* changed and *how much*, not how to reproduce or reverse a change. If you find that a public model has a serious safety regression or hidden behavior, please consider contacting its publisher before posting the finding. See [SECURITY.md](SECURITY.md) for issues in Modeldiffr itself.
+
+## Contributing
+
+Contributions are welcome, from bug reports to new evaluation suites. Start with [CONTRIBUTING.md](CONTRIBUTING.md) and look for issues labelled `good first issue`.
+
+## Citation
+
+If Modeldiffr helps your research, please cite it using [CITATION.cff](CITATION.cff) or:
+
+```bibtex
+@software{hebbar_modeldiffr_2026,
+  author  = {Hebbar, Amaresh},
+  title   = {Modeldiffr: auditing what changes between base and derivative language models},
+  year    = {2026},
+  url     = {https://github.com/amareshhebbar/modeldiffr},
+  version = {0.1.0}
+}
+```
+
+## Acknowledgements
+
+The representational layer will build on crosscoder based model diffing, introduced in *Sparse Crosscoders for Cross Layer Features and Model Diffing* (Lindsey et al., Anthropic, 2024), and on the follow up research it inspired. Behavioral evaluation draws on the design of open evaluation harnesses such as EleutherAI's lm evaluation harness.
+
+## License
+
+[Apache 2.0](LICENSE). Copyright 2026 Amaresh Hebbar.

modeldiffr-0.1.0/examples/README.md

@@ -0,0 +1,18 @@
+# Examples
+
+## Instruct model compared with its base
+
+```bash
+pip install "modeldiffr[hf]"
+python examples/quick_audit.py Qwen/Qwen2.5-0.5B-Instruct Qwen/Qwen2.5-0.5B
+```
+
+This pair is small enough to run on a laptop CPU in a few minutes, and it shows what an instruction tuning delta looks like: a large output divergence, usually with a shift in refusal behavior.
+
+## Sanity check: a model compared with itself
+
+```bash
+modeldiffr audit Qwen/Qwen2.5-0.5B-Instruct Qwen/Qwen2.5-0.5B-Instruct
+```
+
+Every metric should read "within noise". If it does not, something in the setup is not deterministic, and that is worth a bug report.