Oral Session

Oral 1B Datasets & Benchmarks

La Nouvelle Orleans Ballroom A-C (level 2)
Tue 12 Dec 8 a.m. PST — 8:45 a.m. PST

Abstract:

Tue 12 Dec. 8:00 - 8:15 PST

LeanDojo: Theorem Proving with Retrieval-Augmented Language Models

Kaiyu Yang · Aidan Swope · Alex Gu · Rahul Chalamala · Peiyang Song · Shixing Yu · Saad Godil · Ryan J Prenger · Animashree Anandkumar

Large language models (LLMs) have shown promise in proving formal theorems using proof assistants such as Lean. However, existing methods are difficult to reproduce or build on, due to private code, data, and large compute requirements. This has created substantial barriers to research on machine learning methods for theorem proving. This paper removes these barriers by introducing LeanDojo: an open-source Lean playground consisting of toolkits, data, models, and benchmarks. LeanDojo extracts data from Lean and enables interaction with the proof environment programmatically. It contains fine-grained annotations of premises in proofs, providing valuable data for premise selection—a key bottleneck in theorem proving. Using this data, we develop ReProver (Retrieval-Augmented Prover): an LLM-based prover augmented with retrieval for selecting premises from a vast math library. It is inexpensive and needs only one GPU week of training. Our retriever leverages LeanDojo's program analysis capability to identify accessible premises and hard negative examples, which makes retrieval much more effective. Furthermore, we construct a new benchmark consisting of 98,734 theorems and proofs extracted from Lean's math library. It features challenging data split requiring the prover to generalize to theorems relying on novel premises that are never used in training. We use this benchmark for training and evaluation, and experimental results demonstrate the effectiveness of ReProver over non-retrieval baselines and GPT-4. We thus provide the first set of open-source LLM-based theorem provers without any proprietary datasets and release it under a permissive MIT license to facilitate further research.

Tue 12 Dec. 8:15 - 8:30 PST

OpenAssistant Conversations - Democratizing Large Language Model Alignment

Andreas Köpf · Yannic Kilcher · Dimitri von Rütte · Sotiris Anagnostidis · Zhi Rui Tam · Keith Stevens · Abdullah Barhoum · Duc Nguyen · Oliver Stanley · Richárd Nagyfi · Shahul ES · Sameer Suri · David Glushkov · Arnav Dantuluri · Andrew Maguire · Christoph Schuhmann · Huu Nguyen · Alexander Mattick

Aligning large language models (LLMs) with human preferences has proven to drastically improve usability and has driven rapid adoption as demonstrated by ChatGPT.Alignment techniques such as supervised fine-tuning (\textit{SFT}) and reinforcement learning from human feedback (\textit{RLHF}) greatly reduce the required skill and domain knowledge to effectively harness the capabilities of LLMs, increasing their accessibility and utility across various domains.However, state-of-the-art alignment techniques like \textit{RLHF} rely on high-quality human feedback data, which is expensive to create and often remains proprietary.In an effort to democratize research on large-scale alignment, we release OpenAssistant Conversations, a human-generated, human-annotated assistant-style conversation corpus consisting of 161,443 messages in 35 different languages, annotated with 461,292 quality ratings, resulting in over 10,000 complete and fully annotated conversation trees.The corpus is a product of a worldwide crowd-sourcing effort involving over 13,500 volunteers.Models trained on OpenAssistant Conversations show consistent improvements on standard benchmarks over respective base models.We release our code\footnote{\git} and data\footnote{\data} under a fully permissive licence.

Tue 12 Dec. 8:30 - 8:45 PST

DecodingTrust: A Comprehensive Assessment of Trustworthiness in GPT Models

Boxin Wang · Weixin Chen · Hengzhi Pei · Chulin Xie · Mintong Kang · Chenhui Zhang · Chejian Xu · Zidi Xiong · Ritik Dutta · Rylan Schaeffer · Sang Truong · Simran Arora · Mantas Mazeika · Dan Hendrycks · Zinan Lin · Yu Cheng · Sanmi Koyejo · Dawn Song · Bo Li

Generative Pre-trained Transformer (GPT) models have exhibited exciting progress in capabilities, capturing the interest of practitioners and the public alike. Yet, while the literature on the trustworthiness of GPT models remains limited, practitioners have proposed employing capable GPT models for sensitive applications to healthcare and finance – where mistakes can be costly. To this end, this work proposes a comprehensive trustworthiness evaluation for large language models with a focus on GPT-4 and GPT-3.5, considering diverse perspectives – including toxicity, stereotype bias, adversarial robustness, out-of-distribution robustness, robustness on adversarial demonstrations, privacy, machine ethics, and fairness. Based on our evaluations, we discover previously unpublished vulnerabilities to trustworthiness threats. For instance, we find that GPT models can be easily misled to generate toxic and biased outputs and leak private information in both training data and conversation history. We also find that although GPT-4 is usually more trustworthy than GPT-3.5 on standard benchmarks, GPT-4 is more vulnerable given jailbreaking system or user prompts, potentially due to the reason that GPT-4 follows the (misleading) instructions more precisely. Our work illustrates a comprehensive trustworthiness evaluation of GPT models and sheds light on the trustworthiness gaps. Our benchmark is publicly available at https://decodingtrust.github.io/.