Papers by Oliver Li
Beyond Perplexity: Multi-dimensional Safety Evaluation of LLM Compression (2024.findings-emnlp)
Copied to clipboard
| Challenge: | Prior work on compression prioritizes preserving perplexity, which is analogous to training loss. |
| Approach: | They examine the impact of model compression along four dimensions: degeneration harm, representational harm, dialect bias, and language modeling and downstream task performance. |
| Outcome: | The proposed compression methods can lead to unexpected consequences, the authors show . quantization preserves bias while pruning degrades quickly. |
NormDial: A Comparable Bilingual Synthetic Dialog Dataset for Modeling Social Norm Adherence and Violation (2023.emnlp-main)
Copied to clipboard
| Challenge: | Social norms fundamentally shape interpersonal communication. |
| Approach: | They propose a human-in-the-loop pipeline to synthesize a bilingual dyadic dialogue dataset with turn-by-turn annotations of social norms for Chinese and American cultures. |
| Outcome: | The proposed dataset is high-quality through human evaluation and compares with existing models. |
Sociocultural Norm Similarities and Differences via Situational Alignment and Explainable Textual Entailment (2023.emnlp-main)
Copied to clipboard
| Challenge: | Current research on developing computational models of social norms has focused on American society. |
| Approach: | They propose to leverage a Chinese Q&A platform and a socialchiemistry dataset as proxies for contrasting cultural axes and align social situations cross-culturally. |
| Outcome: | The proposed model can reason across cultures using a Chinese Q&A platform and the existing socialChemistry dataset. |
Affective Idiosyncratic Responses to Music (2022.emnlp-main)
Copied to clipboard
| Challenge: | Affective responses to music are highly personal, but it's difficult to measure marginal effects of these variables . a study of 403M listener comments on a social music platform in china aims to address this gap . |
| Approach: | They propose to measure affective responses to music from 403M listener comments on a Chinese social music platform. |
| Outcome: | The proposed method identifies musical, lyrical, contextual, demographic, and mental health effects that drive listener affective responses from over 403M listener comments on a Chinese social music platform. |
How effective is machine translation on low-resource code-switching? A case study comparing human and automatic metrics (2023.findings-acl)
Copied to clipboard
| Challenge: | Specifically, we compare the performance of three MT systems in terms of their ability to translate monolingual Vietnamese, a low-resource language, and Vietnamese-English CSW respectively. |
| Approach: | They compare the performance of three machine translation systems in the context of machine translation (MT) they find that state-of-the-art neural translation systems achieve higher scores on automatic metrics when processing CSW input . |
| Outcome: | The proposed system can translate monolingual Vietnamese, a low-resource language, and Vietnamese-English CSW respectively. |
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. |