--- title: "Diyi Yang — papers, page 2 · Stanford University" description: "Diyi Yang, Stanford University: 48 papers on Elamur Research (2015–2026), h-index 49 — affiliations, topics, co-authors and who cites them…" canonical: "https://research.elamur.ai/free/en/authors/diyi-yang/page/2/" --- [Elamur Research](https://md.research.elamur.ai/free/en/) · [this page in HTML](https://research.elamur.ai/free/en/authors/diyi-yang/page/2/) [home](https://md.research.elamur.ai/free/en/) / [authors](https://md.research.elamur.ai/free/en/authors/) / [Diyi Yang](https://md.research.elamur.ai/free/en/authors/diyi-yang/) / page 2 # Diyi Yang · page 2 48 papers in our library · 2015–2026 · page 2 of 2 [The profile, affiliations and co-authors →](https://md.research.elamur.ai/free/en/authors/diyi-yang/) ## Papers The most cited first (OpenAlex and our library). [← prev](https://md.research.elamur.ai/free/en/authors/diyi-yang/) [1](https://md.research.elamur.ai/free/en/authors/diyi-yang/) **2** 26. **[NormBank: A Knowledge Bank of Situational Social Norms](https://md.research.elamur.ai/free/en/papers/ziems-2023-normbank-knowledge-bank-situational-social/)** *(2023 · 69 citations)* AI assistants, dialogue agents and social robots need to know which behaviours are appropriate. Existing norm resources and reasoners such as Delphi judge actions without context and quietly assume a typical scenario. NormBank addresses this by linking every norm to an explicit situation. 27. **[VALUE: Understanding Dialect Disparity in NLU](https://md.research.elamur.ai/free/en/papers/ziems-2022-value-understanding-dialect-disparity-nlu/)** *(2022 · 67 citations)* The paper builds VALUE, a version of the GLUE benchmark transformed into synthetic African American Vernacular English (AAVE). 28. **[Culturally Aware Natural Language Inference](https://md.research.elamur.ai/free/en/papers/huang-2023-culturally-aware-natural-language-inference/)** *(2023 · 61 citations)* How an utterance is interpreted often depends on cultural norms the listener shares with the speaker. For example, skipping a tip signals displeasure only where tipping is expected. Earlier NLP work mostly catalogued cultural knowledge without testing how that knowledge shapes interpretation. The authors define… 29. **[Unintended Impacts of LLM Alignment on Global Representation](https://md.research.elamur.ai/free/en/papers/ryan-2024-unintended-impacts-llm-alignment-global/)** *(2024 · 57 citations)* Before chat LLMs are released, they are aligned to human preferences through supervised fine-tuning (SFT) and preference tuning (RLHF/PPO or DPO). 30. **[Parameter-Efficient Fine-Tuning Design Spaces](https://md.research.elamur.ai/free/en/papers/chen-2023-parameter-efficient-fine-tuning-design/)** *(2023 · 47 citations)* Existing parameter-efficient fine-tuning (PEFT) techniques such as Adapter, Prefix tuning, BitFit and LoRA were each designed by hand and usually applied identically to every layer. 31. **[SEQZERO: Few-shot Compositional Semantic Parsing with Sequential Prompts and Zero-shot Models](https://md.research.elamur.ai/free/en/papers/yang-2022-seqzero-few-shot-compositional-semantic/)** *(2022 · 38 citations)* The paper tackles semantic parsing (mapping questions to formal queries such as SQL) when only a handful of annotated pairs exist. 32. **[Mitigating Racial Biases in Toxic Language Detection with an Equity-Based Ensemble Framework](https://md.research.elamur.ai/free/en/papers/halevy-2021-mitigating-racial-biases-toxic-language/)** *(2021 · 38 citations)* Toxic-language classifiers trained on Twitter data tend to flag African American English (AAE) posts as offensive more often than Standard American English (SAE) posts. 33. **[Design2Code: Benchmarking Multimodal Code Generation for Automated Front-End Engineering](https://md.research.elamur.ai/free/en/papers/si-2025-design2code-benchmarking-multimodal-code-generation/)** *(2025 · 37 citations)* The paper asks how well multimodal large language models (MLLMs) can turn a webpage screenshot into HTML/CSS code that renders the same page. 34. **[Attacking Vision-Language Computer Agents via Pop-ups](https://md.research.elamur.ai/free/en/papers/zhang-2025-attacking-vision-language-computer-agents/)** *(2025 · 32 citations)* The paper examines a simple but overlooked risk for computer-use agents built on vision-language models (VLMs): visible adversarial pop-ups that an experienced person would dismiss. 35. **[Modeling Gender and Dialect Bias in Automatic Speech Recognition](https://md.research.elamur.ai/free/en/papers/harris-2024-modeling-gender-dialect-bias-automatic/)** *(2024 · 31 citations)* The paper asks whether automatic speech recognition (ASR) errors fall unevenly across speakers when gender and English dialect are considered together, rather than one at a time. 36. **[Modeling Cross-Cultural Pragmatic Inference with Codenames Duet](https://md.research.elamur.ai/free/en/papers/shaikh-2023-modeling-cross-cultural-pragmatic-inference/)** *(2023 · 23 citations)* The paper argues that pragmatic reference, meaning how people work out what a speaker intends, depends on who is talking. 37. **[DADA: Dialect Adaptation via Dynamic Aggregation of Linguistic Rules](https://md.research.elamur.ai/free/en/papers/liu-2023-dada-dialect-adaptation-dynamic-aggregation/)** *(2023 · 20 citations)* DADA adapts a model trained on Standard American English to English dialects without a dialect classifier. 38. **[Distilling an End-to-End Voice Assistant Without Instruction Training Data](https://md.research.elamur.ai/free/en/papers/held-2025-distilling-end-end-voice-assistant/)** *(2025 · 18 citations)* End-to-end Speech LLMs are usually built with supervised finetuning on speech instruction data. That data is scarce, often recorded by few speakers, and the resulting models tend to lose abilities inherited from their text-only backbone. The authors propose the Distilled Voice Assistant (DiVA), which needs no… 39. **[DARG: Dynamic Evaluation of Large Language Models via Adaptive Reasoning Graph](https://md.research.elamur.ai/free/en/papers/zhang-2024-darg-dynamic-evaluation-large-language/)** *(2024 · 17 citations)* Static LLM benchmarks have two weaknesses. Their test items may have leaked into training corpora, and their difficulty stays fixed while models keep improving. Template generators such as DyVal produce rigid, unnatural problems. Prompt-based rewriting (DyVal 2, Benchmark Self-Evolving) offers little control and… 40. **[TADA : Task Agnostic Dialect Adapters for English](https://md.research.elamur.ai/free/en/papers/held-2023-tada-task-agnostic-dialect-adapters/)** *(2023 · 17 citations)* Encoder models such as RoBERTa perform worse when the input is in an English dialect other than Standard American English (SAE). 41. **[Task-Agnostic Low-Rank Adapters for Unseen English Dialects](https://md.research.elamur.ai/free/en/papers/xiao-2023-task-agnostic-low-rank-adapters/)** *(2023 · 13 citations)* Language models trained mostly on Standard American English (SAE) perform noticeably worse for speakers of other English dialects. 42. **[Socially Aware Language Technologies: Perspectives and Practices](https://md.research.elamur.ai/free/en/papers/yang-2025-socially-aware-language-technologies-perspectives/)** *(2025 · 12 citations)* This perspective essay from the "Last Words" column of Computational Linguistics argues that many recurring weaknesses of modern NLP, including biased outputs, toxicity, fragile user trust, uneven performance across groups and difficult evaluation, have a single root. 43. **[Perceptions of Language Technology Failures from South Asian English Speakers](https://md.research.elamur.ai/free/en/papers/holt-2024-perceptions-language-technology-failures-south/)** *(2024 · 9 citations)* NLP systems are known to perform worse on English dialects other than Standard American English (SAmE). It was less clear whether this gap affects real users. The authors study South Asian English (SAsE) speakers living in the US. They ran a Prolific survey with SAsE speakers and a SAmE control group. Participants… 44. **[EquiBench: Benchmarking Large Language Models' Reasoning about Program Semantics via Equivalence Checking](https://md.research.elamur.ai/free/en/papers/wei-2025-equibench-benchmarking-large-language-models/)** *(2025 · 8 citations)* EquiBench asks whether large language models actually grasp what code does or mostly react to how it looks. 45. **[Towards Execution-Grounded Automated AI Research](https://md.research.elamur.ai/free/en/papers/si-2026-towards-execution-grounded-automated-ai/)** *(2026 · 7 citations)* LLMs can propose research ideas that sound convincing but fail when tested. This paper asks two questions: can idea testing be automated at scale, and can models improve from the measured results? The authors built an automated executor. An Implementer turns natural-language ideas into code diffs, a Scheduler assigns… 46. **[AudioJudge: Understanding What Works in Large Audio Model Based Speech Evaluation](https://md.research.elamur.ai/free/en/papers/manakul-2026-audiojudge-understanding-what-works-large/)** *(2026 · 3 citations)* Evaluating generated speech usually requires a separate trained model for each property, such as quality, speaker identity or prosody. 47. **[Future of Work in the Age of LLMs](https://md.research.elamur.ai/free/en/papers/wang-2026-future-work-age-llms/)** *(2026)* This is a two-page tutorial abstract, not an empirical paper, so it presents no new data, experiments, or results. 48. **[Proceedings of the Fourth Workshop on Bridging Human-Computer Interaction and Natural Language Processing (HCI+NLP)](https://md.research.elamur.ai/free/en/papers/blodgett-2025-proceedings-fourth-workshop-bridging-human/)** *(2025)* These are the proceedings of the fourth HCI+NLP workshop. The event took place on 9 November 2025 as part of EMNLP 2025. Its goal is to bring two communities closer together: natural language processing (NLP) and human-computer interaction (HCI). For NLP researchers, the workshop shows how HCI methods help evaluate… [← prev](https://md.research.elamur.ai/free/en/authors/diyi-yang/) [1](https://md.research.elamur.ai/free/en/authors/diyi-yang/) **2**