Papers by Ethan Perez
Red Teaming Language Models with Language Models (2022.emnlp-main)
Copied to clipboard
Ethan Perez, Saffron Huang, Francis Song, Trevor Cai, Roman Ring, John Aslanides, Amelia Glaese, Nat McAleese, Geoffrey Irving
| Challenge: | Prior work has found that language models (LMs) can harm users in hard-to-predict ways, and human annotation is expensive, limiting the number and diversity of test cases. |
| Approach: | They propose to generate test inputs using an LM itself, and use a classifier to detect harmful behavior on test input. |
| Outcome: | The proposed approach detects tens of thousands of offensive responses in a 280B parameter LM chatbot. |
Finding Generalizable Evidence by Learning to Convince Q&A Models (D19-1)
Copied to clipboard
| Challenge: | a system that finds the strongest supporting evidence for a given answer is proposed . a study using passage-based question-answering (QA) shows that agents select evidence that generalizes . |
| Approach: | They propose a system that finds the strongest supporting evidence for a given answer . they use passage-based question-answering (QA) as a testbed to train evidence agents . |
| Outcome: | The proposed system improves QA in a robust manner by using agent-selected evidence. |
Case-based Reasoning for Natural Language Queries over Knowledge Bases (2021.emnlp-main)
Copied to clipboard
Rajarshi Das, Manzil Zaheer, Dung Thai, Ameya Godbole, Ethan Perez, Jay Yoon Lee, Lizhen Tan, Lazaros Polymenakos, Andrew McCallum
| Challenge: | Using human-labeled examples, case-based reasoning can solve complex problems from scratch . case-Based reasoning is a paradigm that is used to solve complex problem . |
| Approach: | They propose a neuro-symbolic CBR approach for question answering over large knowledge bases. |
| Outcome: | The proposed approach outperforms the current state of the art on a CWQ dataset by 11% on accuracy. |
Few-shot Adaptation Works with UnpredicTable Data (2023.acl-long)
Copied to clipboard
| Challenge: | Prior work on language models (LMs) shows that training on a large number of diverse tasks improves few-shot learning (FSL) performance on new tasks. |
| Approach: | They finetuned 413,299 tasks from internet tables to find narrow subsets outperform more diverse datasets. |
| Outcome: | The proposed model outperforms training on 40 human-curated NLP datasets on 52 downstream tasks, but not proportionally to dataset scale. |
RL with KL penalties is better viewed as Bayesian inference (2022.findings-emnlp)
Copied to clipboard
| Challenge: | Reinforcement learning (RL) is used in fine-tuning large language models to penalize them for undesirable features of generated sequences. |
| Approach: | They analyze challenges associated with treating a language model as an RL policy . they find that RL is equivalent to variational inference: approximating a Bayesian posterior . |
| Outcome: | The proposed approach is flawed because it turns the LM into a degenerate distribution, the authors show . they show that the proposed approach avoids the distribution collapse problem and offers a first-principles derivation for its objective. |
Unsupervised Question Decomposition for Question Answering (2020.emnlp-main)
Copied to clipboard
| Challenge: | Existing QA systems struggle to answer complex questions because information is scattered in different places. |
| Approach: | They propose an unsupervised algorithm that decomposes hard questions into simpler sub-questions . they propose an algorithm that can be used to generate a final answer from millions of questions . |
| Outcome: | The proposed algorithm decomposes hard questions into simpler sub-questions that existing QA systems can answer. |
Discovering Language Model Behaviors with Model-Written Evaluations (2023.findings-acl)
Copied to clipboard
Ethan Perez, Sam Ringer, Kamile Lukosiute, Karina Nguyen, Edwin Chen, Scott Heiner, Craig Pettit, Catherine Olsson, Sandipan Kundu, Saurav Kadavath, Andy Jones, Anna Chen, Benjamin Mann, Brian Israel, Bryan Seethor, Cameron McKinnon, Christopher Olah, Da Yan, Daniela Amodei, Dario Amodei, Dawn Drain, Dustin Li, Eli Tran-Johnson, Guro Khundadze, Jackson Kernion, James Landis, Jamie Kerr, Jared Mueller, Jeeyoon Hyun, Joshua Landau, Kamal Ndousse, Landon Goldberg, Liane Lovitt, Martin Lucas, Michael Sellitto, Miranda Zhang, Neerav Kingsland, Nelson Elhage, Nicholas Joseph, Noemi Mercado, Nova DasSarma, Oliver Rausch, Robin Larson, Sam McCandlish, Scott Johnston, Shauna Kravec, Sheer El Showk, Tamera Lanham, Timothy Telleen-Lawton, Tom Brown, Tom Henighan, Tristan Hume, Yuntao Bai, Zac Hatfield-Dodds, Jack Clark, Samuel R. Bowman, Amanda Askell, Roger Grosse, Danny Hernandez, Deep Ganguli, Evan Hubinger, Nicholas Schiefer, Jared Kaplan
| Challenge: | Prior work creates evaluations with crowdwork or existing data sources, which are not always available. |
| Approach: | They generate evaluations automatically with language models (LMs) using crowdwork or existing data sources to find out how they behave . |
| Outcome: | The results show that large LMs repeat back a dialog user’s preferred answer and express greater desire to pursue concerning goals like resource acquisition and goal preservation. |
ELI5: Long Form Question Answering (P19-1)
Copied to clipboard
| Challenge: | Existing question answering datasets provide extractive or short answers, but less attention has been paid to open-ended questions that require explanations. |
| Approach: | They present a large-scale corpus for long form question answering . they use a Reddit forum to provide elaborate answers to open-ended questions . |
| Outcome: | The proposed model outperforms Seq2Seq, language modeling, and other models in human evaluations. |