An official website of the United States government
We present an approach for estimating the fraction of text in a large corpus which is likely to be substantially modified or produced by a large language model (LLM). Our maximum likelihood model leverages expert-written and AI-generated reference texts to accurately and efficiently examine real-world LLM-use at the corpus level. We apply this approach to a case study of scientific peer review in AI conferences that took place after the release of ChatGPT: ICLR 2024, NeurIPS 2023, CoRL 2023 and EMNLP 2023. Our results suggest that between 6.5% and 16.9% of text submitted as peer reviews to these conferences could have been substantially modified by LLMs, i.e. beyond spell-checking or minor writing updates. The circumstances in which generated text occurs offer insight into user behavior: the estimated fraction of LLM-generated text is higher in reviews which report lower confidence, were submitted close to the deadline, and from reviewers who are less likely to respond to author rebuttals. We also observe corpus-level trends in generated text which may be too subtle to detect at the individual level, and discuss the implications of such trends on peer review. We call for future interdisciplinary work to examine how LLM use is changing our information and knowledge practices.
more »
« less
Language Model Sentence Completion with a Parser-Driven Rhetorical Control Method
Controlled text generation (CTG) seeks to guide large language model (LLM) output to produce text that conforms to desired criteria. The current study presents a novel CTG al- gorithm that enforces adherence toward spe- cific rhetorical relations in an LLM sentence- completion context by a parser-driven decoding scheme that requires no model fine-tuning. The method is validated both with automatic and human evaluation. The code is accessible on GitHub.
more »
« less
- Award ID(s):
- 2050919
- PAR ID:
- 10519444
- Publisher / Repository:
- aclanthology.org
- Date Published:
- Format(s):
- Medium: X
- Location:
- Hotel Radisson Blu, St. Julians, in Malta
- Sponsoring Org:
- National Science Foundation
More Like this
-
-
Abstract The prediction of crystal properties plays a crucial role in materials science and applications. Current methods for predicting crystal properties focus on modeling crystal structures using graph neural networks (GNNs). However, accurately modeling the complex interactions between atoms and molecules within a crystal remains a challenge. Surprisingly, predicting crystal properties from crystal text descriptions is understudied, despite the rich information and expressiveness that text data offer. In this paper, we develop and make public a benchmark dataset (TextEdge) that contains crystal text descriptions with their properties. We then propose LLM-Prop, a method that leverages the general-purpose learning capabilities of large language models (LLMs) to predict properties of crystals from their text descriptions. LLM-Prop outperforms the current state-of-the-art GNN-based methods by approximately 8% on predicting band gap, 3% on classifying whether the band gap is direct or indirect, and 65% on predicting unit cell volume, and yields comparable performance on predicting formation energy per atom, energy per atom, and energy above hull. LLM-Prop also outperforms the fine-tuned MatBERT, a domain-specific pre-trained BERT model, despite having 3 times fewer parameters. We further fine-tune the LLM-Prop model directly on CIF files and condensed structure information generated by Robocrystallographer and found that LLM-Prop fine-tuned on text descriptions provides a better performance on average. Our empirical results highlight the importance of having a natural language input to LLMs to accurately predict crystal properties and the current inability of GNNs to capture information pertaining to space group symmetry and Wyckoff sites for accurate crystal property prediction.more » « less
-
Expanded CAG/CTG repeats are sites of DNA damage, leading to changes in repeat length. To determine how ssDNA gap filling affects repeat instability, we inserted (CAG)70 or (CTG)70 repeats into a single-strand annealing (SSA) assay system such that resection and filling in the ssDNA gap would occur across the repeat tract. After resection, when the CTG sequence was the single-stranded template for fill-in synthesis, repeat contractions were elevated and the ssDNA created a fragile site that led to large deletions involving flanking homologous sequences. In contrast, resection was inhibited when CTG was on the resected strand, resulting in repeat expansions. Deleting Rad9, the ortholog of 53BP1, rescued repeat instability and lost viability by increasing resection and fill-in speed. Deletion of Rad51 increased CTG contractions and decreased survival, implicating Rad51 in protecting ssDNA during gap filling. Taken together, DNA sequence within a single-stranded gap determines repair kinetics, fragility, and repeat instability.more » « less
-
Large language models (LLMs) have led to a surge in collaborative writing with model assistance. As different users incorporate suggestions from the same model, there is a risk of decreased diversity in the produced content, potentially limiting diverse perspectives in public discourse. In this work, we measure the impact of co-writing on diversity via a controlled experiment, where users write argumentative essays in three setups -- using a base LLM (GPT3), a feedback-tuned LLM (InstructGPT), and writing without model help. We develop a set of diversity metrics and find that writing with InstructGPT (but not the GPT3) results in a statistically significant reduction in diversity. Specifically, it increases the similarity between the writings of different authors and reduces the overall lexical and content diversity. We additionally find that this effect is mainly attributable to InstructGPT contributing less diverse text to co-written essays. In contrast, the user-contributed text remains unaffected by model collaboration. This suggests that the recent improvement in generation quality from adapting models to human feedback might come at the cost of more homogeneous and less diverse content.more » « less
-
Visual Question Answering (VQA) is a fundamental task in computer vision and natural language process fields. Although the “pre-training & finetuning” learning paradigm significantly improves the VQA performance, the adversarial robustness of such a learning paradigm has not been explored. In this paper, we delve into a new problem: using a pre-trained multimodal source model to create adversarial image-text pairs and then transferring them to attack the target VQA models. Correspondingly, we propose a novel VQATTACK model, which can iteratively generate both im- age and text perturbations with the designed modules: the large language model (LLM)-enhanced image attack and the cross-modal joint attack module. At each iteration, the LLM-enhanced image attack module first optimizes the latent representation-based loss to generate feature-level image perturbations. Then it incorporates an LLM to further enhance the image perturbations by optimizing the designed masked answer anti-recovery loss. The cross-modal joint attack module will be triggered at a specific iteration, which updates the image and text perturbations sequentially. Notably, the text perturbation updates are based on both the learned gradients in the word embedding space and word synonym-based substitution. Experimental results on two VQA datasets with five validated models demonstrate the effectiveness of the proposed VQATTACK in the transferable attack setting, compared with state-of-the-art baselines. This work revealsa significant blind spot in the “pre-training & fine-tuning” paradigm on VQA tasks. The source code can be found in the link https://github.com/ericyinyzy/VQAttack.more » « less
- Free Publicly Accessible Full Text
-
Accepted Manuscript1.0
- Conference Proceeding:
- The DOI is not currently available.
