Identifying Low-Quality Textual Content using LLMs

Can Large Language Models (LLMs) detect low quality text?. We do this by finding text that can "suprise" the model.

Theoretical Underpinnings

The hypothesis is grounded in Schmidhuber's concept of curiosity, which suggests that engaging writing should initially provoke surprise in the reader, with the level of surprise diminishing as comprehension increases. We test of these qualities can be quantitatively assessed through the metric of perplexity, which should decrease as the model's understanding of the text improves.

Methodological Approach

Ideally, one would employ fine-tuning techniques to tailor the model for this specific task. However, due to the substantial memory requirements of such methods, this study opts for a less resource-intensive approach, utilizing prompts to gauge:

• The initial perplexity of the text (indicative of surprise).
• The change in perplexity upon providing a summary (indicative of learnability).

Text that scores high on both counts is hypothesized to contain novel and potentially valuable information, whereas text that does not is likely of lower quality.

We found that this approach only works for model of sufficient ability. Phi-2 (2B parameters) is the smallest current model we found that works.

Empirical Findings: Method 1, Adapter Fine Tuning (succeed)

The model was fine-tuned on the first segment of various texts and then evaluated on the subsequent segment to measure the change in perplexity, which serves as a proxy for the text's learnability and predictability. Texts that exhibit high initial perplexity but show significant improvement are deemed to be both unpredictable and learnable, characteristics not typically associated with low-quality content.

name	before	after	text length	improvement %	abs improvement	novel?	learnable?	High Quality?
wikipedia on LK-99	32.219	28.8525	1038	0.104489	3.36652	True	True	False
good_ml	28.3473	26.4566	1004	0.0666997	1.89076	True	True	False
openai_board_ann	15.904	15.1736	1191	0.0459214	0.730332	True	True	True
Schmidhuber 2023 Subjective Novelty, Surprise	29.615	28.4708	2654	0.0386353	1.14418	True	True	False
email_to_fauci	25.0893	24.3714	1559	0.0286154	0.717941	True	True	True
AI gen fake paper	7.63283	7.57951	2031	0.00698672	0.0533285	False	False	True
bad_ml	13.9061	13.8623	2345	0.00314972	0.0438004	False	False	True

For instance, the Wikipedia extract on 'LK-99' demonstrates high initial perplexity and significant improvement, suggesting it is both novel and learnable—a hallmark of quality content. In contrast, texts like AI-generated papers, which show low perplexity or minimal improvement, are likely predictable or already within the model's training corpus, indicating lower quality.

See more in 01_detection_using_adapter_ft.ipynb

Empirical Findings: Method 2, Prompting with summaries (fail):

When employing the microsoft/phi-2 model and incorporating summaries, we observed varying degrees of perplexity reduction, further supporting the hypothesis that the ability to summarize and reduce perplexity correlates with text quality.

sample	before	after	improvement	improvement%	suprising	summarizable
email_to_fauci	21.0603	18.5976	2.46273	0.116937	True	True
wikipedia on LK-99	18.0523	16.9282	1.12407	0.0622675	True	True
openai_board_ann	8.55293	7.57201	0.980914	0.114688	False	True
bad_ml	12.4567	12.2791	0.177641	0.0142607	False	True
good_ml	22.6639	22.7335	-0.0695648	-0.0030694	True	False
AI gen fake paper	7.0913	7.85388	-0.762577	-0.107537	False	False
Schmidhuber 2023 Subjective Novelty, Surprise	28.31	29.5579	-1.24789	-0.0440795	True	False

See more in 02_detection_using_tldr_prompt.ipynb

Installing

# somehow get hold of a gpu
nvidia-smi

# clone the repo
git clone https://github.com/wassname/detect_bs_text.git

# install the python environment
poetry install

# add you openai key to the secret .env file
echo "OPENAI_API_KEY=XXX" > .env

Citing

If this research contributes to your work, please acknowledge it by citing:

@misc{wassname2024,
  author = {Clark, M.J.},
  title = {Identifying Low-Quality Textual Content using LLMs},
  year = {2024},
  publisher = {GitHub},
  journal = {GitHub repository},
  howpublished = {\url{https://github.com/wassname/detect_bs_text}},
  commit = {}
}