https://github.com/waltonfuture/Diff-eRank
Code for https://arxiv.org/abs/2401.17139 (NeurIPS 2024)
https://github.com/waltonfuture/Diff-eRank
evaluation-metrics llm llm-inference machine-learning mllm neurips-2024
Last synced: 3 months ago
JSON representation
Code for https://arxiv.org/abs/2401.17139 (NeurIPS 2024)
- Host: GitHub
- URL: https://github.com/waltonfuture/Diff-eRank
- Owner: waltonfuture
- License: apache-2.0
- Created: 2024-01-29T13:16:52.000Z (almost 2 years ago)
- Default Branch: master
- Last Pushed: 2024-11-15T11:46:37.000Z (12 months ago)
- Last Synced: 2024-11-15T12:31:54.292Z (12 months ago)
- Topics: evaluation-metrics, llm, llm-inference, machine-learning, mllm, neurips-2024
- Language: Python
- Homepage: https://arxiv.org/abs/2401.17139
- Size: 37.1 KB
- Stars: 23
- Watchers: 1
- Forks: 2
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
- License: LICENSE
Awesome Lists containing this project
- StarryDivineSky - waltonfuture/Diff-eRank - eRank是一种基于排序的指标，用于评估大型语言模型（LLM），它基于信息论和几何原理，通过分析模型的隐藏表示来量化模型在训练后丢弃冗余信息的能力。该指标适用于单模态（语言）和多模态场景。研究发现，Diff-eRank在模型规模扩大时会增加，并且与传统的指标（如损失和准确率）保持一致的关系。该项目提供了代码和示例，可以计算单个句子或数据集的 Diff-eRank，并提供了相应的论文和项目链接。 (A01_文本生成_文本对话 / 大语言对话模型及数据)
README
# Diff-eRank: A Novel Rank-Based Metric for Evaluating Large Language Models (NeurIPS 2024)
[Lai Wei](https://waltonfuture.github.io/) *, Zhiquan Tan *, Chenghai Li, [Jindong Wang](https://jd92.wang/), [Weiran Huang](https://www.weiranhuang.com/) (*Equal Contribution).
**Shanghai Jiao Tong University & Tsinghua University & William and Mary**
## Introduction
We introduce a rank-based metric called Diff-eRank, which is rooted in information theory and geometry principles. Diff-eRank evaluates LLMs by examining their hidden representations to quantify how LLMs discard redundant information after training.
Specifically, we demonstrate its applicability in both single-modal (language) and multi-modal settings. For language models, our findings reveal that the Diff-eRank increases when the model scales up, which also demonstrates a consistent relationship with traditional metrics like loss and accuracy.
For multi-modal models, we also propose an evaluation method based on rank for assessing alignment quality and we find that modern multi-modal large language models exhibit good alignment performance.
## Calculation of Diff-eRank for LLMs
### Setup
```bash
pip install transformers torch datasets
```
### Calculation
```bash
from transformers import AutoTokenizer, AutoModel, AutoConfig
import torch
import math
# R input N*d
def normalize(R):
with torch.no_grad():
mean = R.mean(dim=0)
R = R - mean
norms = torch.norm(R, p=2, dim=1, keepdim=True)
R = R/norms
return R
def cal_cov(R):
with torch.no_grad():
Z = torch.nn.functional.normalize(R, dim=1)
A = torch.matmul(Z.T, Z)/Z.shape[0]
return A
def cal_erank(A):
with torch.no_grad():
eig_val = torch.svd(A / torch.trace(A))[1]
entropy = - (eig_val * torch.log(eig_val)).nansum().item()
erank = math.exp(entropy)
return erank
def compute(R):
return cal_erank(cal_cov(normalize(R)))
model_path = "facebook/opt-1.3b" # for example
tokenizer = AutoTokenizer.from_pretrained(model_path)
model = AutoModel.from_pretrained(model_path).cuda()
config = AutoConfig.from_pretrained(model_path)
untrained_model = AutoModel.from_config(config).to('cuda')
text = "We introduce a rank-based metric called Diff-eRank, which is rooted in information theory and geometry principles. Diff-eRank evaluates LLMs by examining their hidden representations to quantify how LLMs discard redundant information after training." # for example
inputs = tokenizer(text, return_tensors="pt").to('cuda')
with torch.no_grad():
R1 = model(inputs.input_ids)[0][0, :, :]
R2 = untrained_model(inputs.input_ids)[0][0, :, :]
erank1 = compute(R1)
erank2 = compute(R2)
RD = erank2 - erank1
print(RD)
```
### Diff-eRank of Single Sentence
```
cd utils
python diff_erank_single_sentence.py
```
### Diff-eRank of Dataset
Please download the datasets of [wiki-en](https://huggingface.co/datasets/wikipedia), [dolly-15k](https://huggingface.co/datasets/databricks/databricks-dolly-15k), [openwebtext2](https://huggingface.co/datasets/suolyer/pile_openwebtext2), [hh-rlhf](https://huggingface.co/datasets/Anthropic/hh-rlhf) in huggingface and edit the data path in your scripts.
```
cd utils
python diff_erank_dataset.py
```
## Calculate eRank for MLLMs
We provide an example script to calculate eRank for Qwen2.5-VL. Please check it in ```utils/erank-qwen2_5_vl.py```
## Citation
If you're using Diff-eRank in your research or applications, please cite using this BibTeX:
```bibtex
@inproceedings{weidiff,
title={Diff-eRank: A Novel Rank-Based Metric for Evaluating Large Language Models},
author={Wei, Lai and Tan, Zhiquan and Li, Chenghai and Wang, Jindong and Huang, Weiran},
booktitle={The Thirty-eighth Annual Conference on Neural Information Processing Systems},
year={2024}
}
```