Metadata-Version: 2.4
Name: llmcostcut
Version: 1.0.2
Summary: A framework for selectively invoking LLMs and distilling repeated workloads into smaller models
Home-page: https://github.com/zhaoliangvaio/llmcostcut
Author: Liang Zhao
Author-email: Liang Zhao <liang.zhao@emory.edu>, LLMCostCut Team <liang.zhao@emory.edu>
License: 
                                         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
        
           APPENDIX: How to apply the Apache License to your work.
        
              To apply the Apache License to your work, attach the following
              boilerplate notice, with the fields enclosed by brackets "[]"
              replaced with your own identifying information. (Don't include
              the brackets!)  The text should be enclosed in the appropriate
              comment syntax for the file format. We also recommend that a
              file or class name and description of purpose be included on the
              same "printed page" as the copyright notice for easier
              identification within third-party archives.
        
           Copyright [yyyy] [name of copyright owner]
        
           Licensed under the Apache License, Version 2.0 (the "License");
           you may not use this file except in compliance with the License.
           You may obtain a copy of the License at
        
               http://www.apache.org/licenses/LICENSE-2.0
        
           Unless required by applicable law or agreed to in writing, software
           distributed under the License is distributed on an "AS IS" BASIS,
           WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
           See the License for the specific language governing permissions and
           limitations under the License.
        
Project-URL: Homepage, https://github.com/zhaoliangvaio/llmcostcut
Project-URL: Documentation, https://github.com/zhaoliangvaio/llmcostcut#readme
Project-URL: Repository, https://github.com/zhaoliangvaio/llmcostcut
Project-URL: Issues, https://github.com/zhaoliangvaio/llmcostcut/issues
Keywords: llm,distillation,knowledge-distillation,online-learning,adaptive-inference,machine-learning
Classifier: Development Status :: 3 - Alpha
Classifier: Intended Audience :: Developers
Classifier: Intended Audience :: Science/Research
Classifier: License :: OSI Approved :: Apache Software License
Classifier: Programming Language :: Python :: 3
Classifier: Programming Language :: Python :: 3.8
Classifier: Programming Language :: Python :: 3.9
Classifier: Programming Language :: Python :: 3.10
Classifier: Programming Language :: Python :: 3.11
Classifier: Programming Language :: Python :: 3.12
Classifier: Topic :: Scientific/Engineering :: Artificial Intelligence
Classifier: Topic :: Software Development :: Libraries :: Python Modules
Requires-Python: >=3.8
Description-Content-Type: text/markdown
License-File: LICENSE
Requires-Dist: torch>=1.9.0
Requires-Dist: transformers>=4.20.0
Requires-Dist: datasets>=2.0.0
Requires-Dist: openai>=1.0.0
Requires-Dist: python-dotenv>=1.0.0
Provides-Extra: dev
Requires-Dist: pytest>=7.0.0; extra == "dev"
Requires-Dist: pytest-cov>=4.0.0; extra == "dev"
Requires-Dist: black>=22.0.0; extra == "dev"
Requires-Dist: flake8>=5.0.0; extra == "dev"
Requires-Dist: mypy>=1.0.0; extra == "dev"
Dynamic: author
Dynamic: home-page
Dynamic: license-file
Dynamic: requires-python


<p align="center">
  <img src="https://raw.githubusercontent.com/zhaoliangvaio/llmcostcut/main/assets/llmcostcut_logo.png" alt="LLMCostCut Logo" width="600"/>
</p>

# LLMCostCut: The Intelligent Cost Cutter for Large Language Models

**Optimized. Efficient. Powerful.**

[![GitHub](https://img.shields.io/badge/GitHub-Project-black?logo=github)](https://github.com/zhaoliangvaio/llmcostcut)
[![arXiv](https://img.shields.io/badge/arXiv-2602.03006-b31b1b?logo=arxiv)](https://arxiv.org/abs/2602.03006)
[![License](https://img.shields.io/badge/License-MIT-green)](https://raw.githubusercontent.com/zhaoliangvaio/llmcostcut/main/LICENSE)
[![Python](https://img.shields.io/badge/Python-3.8-blue?logo=python)](https://www.python.org/)
[![PyPI](https://img.shields.io/badge/PyPI-v1.0.2-orange?logo=pypi)](https://pypi.org/project/llmcostcut/)


---

As LLM becomes more and more popular, the cost of using LLM is becoming a major concern. LLMCostCut is a discriminative workload for LLM to reduce the cost of using LLM while maintaining the accuracy.

## Table of Contents

- 📋 [Applications](#applications)
- 🎯 [Key Features](#key-features)
- 📦 [Installation](#installation)
  - 📋 [Requirements](#requirements)
- 🚀 [Quick Start](#quick-start)
- ⚙️ [How It Works](#how-it-works)
- 📊 [Experimental Results](#experimental-results)
  - 📈 [Accuracy](#accuracy)
  - 💰 [Cost Reduction](#cost-reduction)
  - 📊 [Benchmarks Comparison](#benchmarks-comparison)
- 📁 [Project Structure](#project-structure)
- 📚 [API Reference](#api-reference)
  - 🔧 [monitor()](#monitor---main-parameters)
- 📖 [Citation](#citation)
- 🙏 [Acknowledgements](#acknowledgements)


## 📋 Applications
Reducing Cost of LLM Reasoning in Discriminative Workloads

<p align="center">
  <img src="https://raw.githubusercontent.com/zhaoliangvaio/llmcostcut/main/assets/table.png" alt="Applications" width="900"/>
</p>

<!-- | ⚖️ Legal | 🏥 Healthcare | 💼 Finance | ✨ More |
|---|---|---|---|
| -Case Outcome Prediction<br>-Contract Clause Classification<br>-Legal Reasoning / Multi-hop QA<br><div align="center">...</div> | -Radiology Report Abnormality Classification<br>-Clinical Note Coding<br>-Medical NER / Drug-Drug Interaction<br><div align="center">...</div> | -Sentiment Analysis<br>-Risk Assessment<br>-Fraud Detection / Financial QA<br><div align="center">...</div> | -Science<br>-Materials<br><div align="center">...</div> | -->

## 🎯 Key Features


| 📈 **Accuracy up to 95%**                                         | 💰 **Around 10× Cost Reduction**                                                              | ⚡ **Nearly 1000× Speedup**                                                                                                     | 🛠️ **Easy to Use**                                  |
| ----------------------------------------------------------------- | --------------------------------------------------------------------------------------------- | ------------------------------------------------------------------------------------------------------------------------------ | ---------------------------------------------------- |
| Maintain near-teacher quality while cutting LLM calls toward zero | Student handles most queries; teacher only when uncertain — inference bills drop dramatically | Local tiny student model (~100k-2M params) inference vs. LLM (~1T params) API calls — orders of magnitude faster in latency | Minimal code changes — Runnable in few lines of code |


---

## 📦 Installation

**Install from source** (recommended for development)

```bash
git clone https://github.com/zhaoliangvaio/llmcostcut.git
cd llmcostcut
pip install -e .
```

**Install from PyPI**

```bash
pip install llmcostcut
```

### 📋 Requirements

- Python >= 3.8
- PyTorch >= 1.9.0
- Transformers >= 4.20.0

Optional (for built-in OpenAI teacher and examples):

- `openai>=1.0.0`
- `python-dotenv>=1.0.0`
- `pydantic>=2.0.0`

When using the built-in OpenAI teacher, set your API key:

```bash
export OPENAI_API_KEY="your-key-here"
```

Or put `OPENAI_API_KEY=your-key-here` in a `.env` file in the project root (loaded automatically if `python-dotenv` is installed).

## 🚀 Quick Start

Instead of calling the LLM directly every time, use `**monitor**` as a smart router: it first tries the tiny student model; only when the student is uncertain (confidence below `p_threshold`) does it fall back to the teacher LLM. Over time, the student learns and LLM calls drop toward zero.

```python
from datasets import load_dataset
from llmcostcut.monitor import monitor
from openai import OpenAI

def classify_with_llm(texts, task):
    labels, client, results = task["topic_classification"], OpenAI(), []
    for text in texts:
        r = client.chat.completions.create(
            model="gpt-4o-mini",
            messages=[{"role":"user","content":f"Classify into {labels}. Text: {text[:300]}"}],
            max_tokens=10,
        )
        results.append({"topic_classification": r.choices[0].message.content.strip()})
    return results

TASK = {"topic_classification": ["World","Sports","Business","Sci/Tech"]}

for example in load_dataset("ag_news", split="train[:100]"):
    output, used_llm_or_not = monitor(TASK, example["text"], llm_fn=classify_with_llm, mode="online")

monitor.close()
```

See **[examples/example.py](https://raw.githubusercontent.com/zhaoliangvaio/llmcostcut/main/examples/example.py)** for a full AG-News demo with GCP classifier and concept-level labels.

---

## ⚙️ How It Works

![LLMCostCut](https://raw.githubusercontent.com/zhaoliangvaio/llmcostcut/main/assets/LLMCostCut.gif)

---

## 📊 Experimental Results

### 💰 Cost Reduction

![Cost Curve](https://raw.githubusercontent.com/zhaoliangvaio/llmcostcut/main/assets/cost_curve.png)

The inference cost decreases over time as the student model handles more queries and the fallback to the teacher LLM becomes less frequent.

### 📈 Accuracy

![Accuracy](https://raw.githubusercontent.com/zhaoliangvaio/llmcostcut/main/assets/acc.png)

System accuracy stays close to the teacher baseline (100%) during training. Despite the drop in LLM fallback, accuracy stabilizes around 95% after the initial phase, showing that the distilled student preserves quality while cutting inference cost.

### 📊 Benchmarks Comparison in Accuracy


| Dataset                                       | LLMCost with Multilayer Perceptron (MLP) as student model | LLMCost with Graph of Concepts (GCP) as student model |
| --------------------------------------------- | ------------------------------------------------------- | --------------------------------------------------- |
| Supreme Court Judgment Prediction Dataset     | 95.8 %                                                   | 97.6 %                                               |
| MIMIC-CXR Dataset                             | 93.6 %                                                   | 96.5 %                                               |
| American Express - Default Prediction Dataset | 95.7 %                                                   | 97.8 %                                               |


Across [Supreme Court Judgment Prediction Dataset](https://www.kaggle.com/datasets/deepcontractor/supreme-court-judgment-prediction), [MIMIC-CXR Dataset](https://physionet.org/content/mimic-cxr/2.1.0/), and [American Express - Default Prediction Dataset](https://www.kaggle.com/competitions/amex-default-prediction), our student models **approximately comparable to the teacher LLM baseline** in *relative accuracy*: GCP (proposed in [Distilling LLM Reasoning into Graph of Concept Predictors](https://arxiv.org/abs/2602.03006)) reaches 97.6%, 96.5%, and 97.8% respectively, while MLP achieves 95.8%, 93.6%, and 95.7%. This demonstrates that the framework preserves prediction quality while enabling efficient inference.

---

## 📁 Project Structure

```
llmcostcut/
├── src/
│   └── llmcostcut/
│       __init__.py       # Exports monitor
│       buffers.py       # Replay buffer (RingBuffer, balanced sampling)
│       correctness.py   # Correctness predictor
│       defaults.py      # Encoder, tokenizer, optimizer defaults
│       models.py        # Classifier heads (DeepMLP, GCP)
│       monitor.py       # Core monitor() API and training orchestration
│       registry.py      # Per-task registry
│       selector.py      # Active-learning / offline sample selection
│       task.py          # Task abstraction
│       trainer.py       # Training and GCP submodule retraining
├── examples/
│   example.py           # AG-News with GCP classifier (concept DAG + online/offline)
├── setup.py
├── pyproject.toml
├── requirements.txt
├── README.md
└── LICENSE
```

---

## 📚 API Reference

### 🔧 `monitor(...)` — main parameters

#### Required

- `**tasks**` (`dict[str, list[str]]`): Task ID → list of allowed class labels.
  ```python
  {"task1": ["class1", "class2"], "task2": ["A", "B", "C"]}
  ```
- `**text**` (`str | list[str] | tuple[str, ...]`): Input text(s). Single string for one sample, list/tuple for batch.
- `**mode**` (str): **Required.** `"online"` or `"offline"`.
  - `"online"`: student is updated during inference.
  - `"offline"`: student is frozen; samples are selected up-front via `offline_select_method` and `offline_select_budget`.

#### Optional parameters


| Parameter           | Default                   | Description                                                                                        |
| ------------------- | ------------------------- | -------------------------------------------------------------------------------------------------- |
| `llm_fn`            | built-in OpenAI           | Teacher function. Custom signature: `llm_fn(texts, tasks, **kwargs)`.                              |
| `p_threshold`       | `0.8`                     | Min confidence to trust the student; below this, use LLM.                                          |
| `classifier_type`   | `"deep_mlp"`              | Student head: `"deep_mlp"` or `"gcp"`.                                                             |
| `classifier_kwargs` | `None`                    | Architecture-specific kwargs (see below).                                                          |
| `encoder`           | `distilbert-base-uncased` | Encoder model instance.                                                                            |
| `device`            | auto                      | Device (e.g. `"cuda:0"`, `"cpu"`).                                                                 |
| `llm_kwargs`        | `{}`                      | Passed to `llm_fn`. For default OpenAI teacher, can include `concept_info` for GCP concept labels. |


#### Offline-only parameters


| Parameter               | Description                                                                       |
| ----------------------- | --------------------------------------------------------------------------------- |
| `offline_select_method` | **Required** when `mode="offline"`. Strategy (e.g. `"random"`, `"uncertainty"`).  |
| `offline_select_budget` | **Required** when `mode="offline"`. Max number of samples to send to the teacher. |


#### Return value

A 2-tuple `(results, used_llm)`:


| Input          | `results`              | `used_llm`   |
| -------------- | ---------------------- | ------------ |
| Single `str`   | `dict[str, str]`       | `bool`       |
| list/tuple str | `list[dict[str, str]]` | `list[bool]` |


---

## 📖 Citation

If you use this framework in your research, please cite:

```bibtex
@article{yu2026distilling,
  title     = {Distilling {LLM} Reasoning into Graph of Concept Predictors},
  author    = {Ziyang Yu and Liang Zhao},
  journal   = {arXiv preprint arXiv:2602.03006},
  year      = {2026},
  url       = {https://arxiv.org/abs/2602.03006},
  doi       = {10.48550/arXiv.2602.03006},
}
```

---

## 🙏 Acknowledgements

This framework was developed by the team led by Prof. Liang Zhao at Emory University. We thank collaborators and students for feedback. The design draws on work in online learning, knowledge distillation, and adaptive inference.
