--- title: "Timothy Baldwin — papers, page 2 · The University of Melbourne" description: "Timothy Baldwin, The University of Melbourne: 53 papers on Elamur Research (2002–2026), h-index 54 — affiliations, topics, co-authors and…" canonical: "https://research.elamur.ai/free/en/authors/timothy-baldwin/page/2/" --- [Elamur Research](https://md.research.elamur.ai/free/en/) · [this page in HTML](https://research.elamur.ai/free/en/authors/timothy-baldwin/page/2/) [home](https://md.research.elamur.ai/free/en/) / [authors](https://md.research.elamur.ai/free/en/authors/) / [Timothy Baldwin](https://md.research.elamur.ai/free/en/authors/timothy-baldwin/) / page 2 # Timothy Baldwin · page 2 53 papers in our library · 2002–2026 · page 2 of 3 [The profile, affiliations and co-authors →](https://md.research.elamur.ai/free/en/authors/timothy-baldwin/) ## Papers The most cited first (OpenAlex and our library). [← prev](https://md.research.elamur.ai/free/en/authors/timothy-baldwin/) [1](https://md.research.elamur.ai/free/en/authors/timothy-baldwin/) **2** [3](https://md.research.elamur.ai/free/en/authors/timothy-baldwin/page/3/) [next →](https://md.research.elamur.ai/free/en/authors/timothy-baldwin/page/3/) 26. **[Do-Not-Answer: Evaluating Safeguards in LLMs](https://md.research.elamur.ai/free/en/papers/wang-2024-do-not-answer-evaluating-safeguards/)** *(2024 · 93 citations)* Open-source LLMs often ship without the safety layers that commercial systems have. Existing harmful-prompt collections tend to be small, proprietary, or unlabeled. This paper offers a cheap, public way to test whether a model refuses requests it should not fulfil. 27. **[Multiword expressions: linguistic precision and reusability](https://md.research.elamur.ai/free/en/papers/copestake-2002-multiword-expressions-linguistic-precision-reusability/)** *(2002 · 80 citations)* The paper describes how the LinGO English Resource Grammar (ERG), a broad bidirectional HPSG grammar, encodes multiword expressions (MWEs) in a database. 28. **[Are Multilingual LLMs Culturally-Diverse Reasoners? An Investigation into Multicultural Proverbs and Sayings](https://md.research.elamur.ai/free/en/papers/liu-2023-multilingual-llms-culturally-diverse-reasoners/)** *(2023 · 78 citations)* This paper asks whether multilingual LLMs (mLLMs) can reason with culture-specific shared knowledge, using proverbs and sayings as the test case. 29. **[Do-Not-Answer: A Dataset for Evaluating Safeguards in LLMs](https://md.research.elamur.ai/free/en/papers/wang-2023-answer-dataset-evaluating-safeguards-llms/)** *(2023 · 78 citations)* Open-source LLMs often ship without strong safety mechanisms, and earlier harmful-prompt collections were either private, narrow in scope, or lacked labeled model responses. 30. **[Language models are not naysayers: An analysis of language models on negation benchmarks](https://md.research.elamur.ai/free/en/papers/truong-2023-language-models-naysayers-analysis-language/)** *(2023 · 78 citations)* Earlier research showed that masked models such as BERT handle negation poorly. This paper asks whether that weakness persists in larger auto-regressive LLMs. The authors assembled six English benchmarks in which negation determines the correct answer and grouped them by the skill each one tests: 31. **[Learning the Countability of English Nouns from Corpus Data](https://md.research.elamur.ai/free/en/papers/baldwin-2003-learning-countability-english-nouns-corpus/)** *(2003 · 78 citations)* Baldwin and Bond tackle automatic acquisition of noun countability: deciding, for each English noun type, whether it behaves as countable, uncountable, bipartite (scissors-like) or plural only. 32. **[Benchmarking Uncertainty Quantification Methods for Large Language Models with LM-Polygraph](https://md.research.elamur.ai/free/en/papers/vashurin-2025-benchmarking-uncertainty-quantification-methods-large/)** *(2025 · 67 citations)* Uncertainty quantification (UQ) can help flag hallucinated or low-quality outputs of large language models. Earlier work on UQ for LLMs used incompatible methods and evaluation setups, so findings were hard to compare. 33. **[IndoCulture: Exploring Geographically Influenced Cultural Commonsense Reasoning Across Eleven Indonesian Provinces](https://md.research.elamur.ai/free/en/papers/koto-2024-indoculture-exploring-geographically-influenced-cultural/)** *(2024 · 53 citations)* Most commonsense reasoning benchmarks are built in English and ignore where the reasoner lives, so they tend to reward Anglocentric knowledge. 34. **[Uncertainty Estimation and Reduction of Pre-trained Models for Text Regression](https://md.research.elamur.ai/free/en/papers/wang-2022-uncertainty-estimation-reduction-pre-trained/)** *(2022 · 49 citations)* Pre-trained language models used for text regression tend to be overconfident when they are wrong. This limits their use in high-stakes areas such as clinical decision support. Calibration work in NLP has mostly targeted classifiers, so Wang et al. study how well regression models built on pre-trained LMs can express… 35. **[Top-down Discourse Parsing via Sequence Labelling](https://md.research.elamur.ai/free/en/papers/koto-2021-top-down-discourse-parsing-sequence/)** *(2021 · 37 citations)* The paper addresses document-level discourse parsing under Rhetorical Structure Theory (RST). Most earlier parsers build trees bottom-up, which tends to ignore global document context. Recent top-down parsers have relied on complex machinery such as decoders, stacks or exhaustive split scoring. The authors recast… 36. **[Classifying Dialogue Acts in Multi-party Live Chats](https://md.research.elamur.ai/free/en/papers/kim-2012-classifying-dialogue-acts-multi-party/)** *(2012 · 34 citations)* Kim, Cavedon and Baldwin study how to automatically label chat messages with dialogue acts when many people talk at once. 37. **[What does it take to bake a cake? The RecipeRef corpus and anaphora resolution in procedural text](https://md.research.elamur.ai/free/en/papers/fang-2022-what-does-it-take-bake/)** *(2022 · 32 citations)* Most anaphora research uses declarative genres such as news or dialogue. Procedural text, where readers must track which entity each step acts on and how it changes, has received little attention. The authors address this for cooking recipes. They first compare the textual properties of recipes and chemical patents.… 38. **[Large Language Models Are Human-Like Internally](https://md.research.elamur.ai/free/en/papers/kuribayashi-2025-large-language-models-are-human/)** *(2025 · 30 citations)* Several recent psycholinguistic studies found that surprisal from bigger language models predicts human reading data worse than surprisal from small models such as GPT-2 small. 39. **[Evaluating Document Coherence Modeling](https://md.research.elamur.ai/free/en/papers/shen-2021-evaluating-document-coherence-modeling/)** *(2021 · 30 citations)* The paper asks whether pretrained language models genuinely understand document coherence or only exploit surface cues. 40. **[ChEMU-Ref: A Corpus for Modeling Anaphora Resolution in the Chemical Domain](https://md.research.elamur.ai/free/en/papers/fang-2021-chemu-ref-corpus-modeling-anaphora/)** *(2021 · 27 citations)* Building semi-structured knowledge from chemical patents requires tracking how compounds are referenced and changed across a reaction description. 41. **[Not another Negation Benchmark: The NaN-NLI Test Suite for Sub-clausal Negation](https://md.research.elamur.ai/free/en/papers/truong-2022-not-another-negation-benchmark-nan/)** *(2022 · 24 citations)* Pre-trained language models often mishandle negation, and existing negation benchmarks mostly add a plain not to a main verb. 42. **[A Plethora of Methods for Learning English Countability](https://md.research.elamur.ai/free/en/papers/baldwin-2003-plethora-methods-learning-english-countability/)** *(2003 · 23 citations)* The paper addresses deep lexical acquisition: assigning English nouns to countability classes (countable, uncountable, plural only, bipartite) from corpus evidence. 43. **[Arabic Dataset for LLM Safeguard Evaluation](https://md.research.elamur.ai/free/en/papers/ashraf-2025-arabic-dataset-llm-safeguard-evaluation/)** *(2025 · 22 citations)* Most safety benchmarks for large language models are built for English, with some extensions to Chinese. Arabic, with its cultural and political sensitivities, has had little coverage. The authors build an Arabic safety benchmark in two steps. First, they adapt the general-risk portion of the Chinese CDNA dataset… 44. **[Token-Level Density-Based Uncertainty Quantification Methods for Eliciting Truthfulness of Large Language Models](https://md.research.elamur.ai/free/en/papers/vazhentsev-2025-token-level-density-based-uncertainty/)** *(2025 · 13 citations)* LLMs often produce hallucinated or non-factual output. Many uncertainty quantification (UQ) methods can flag such output, but the strongest ones sample the model repeatedly and are expensive. Density-based scores such as Mahalanobis Distance (MD) are cheap and work well for encoder classifiers, yet earlier… 45. **[Target Based Speech Act Classification in Political Campaign Text](https://md.research.elamur.ai/free/en/papers/subramanian-2019-target-based-speech-act-classification/)** *(2019 · 13 citations)* The paper looks at pragmatics in election campaign material. For each utterance it asks two questions: what the politician is doing (the speech act), and which party the utterance concerns (the target). The authors built a new corpus of press releases and speech transcripts from Labor and Liberal, the two major… 46. **[Automatic Negation and Speculation Detection in Veterinary Clinical Text](https://md.research.elamur.ai/free/en/papers/cheng-2017-automatic-negation-speculation-detection-veterinary/)** *(2017 · 11 citations)* The paper tackles negation and speculation detection in veterinary clinical notes from the VetCompass project. 47. **[Unconditional Truthfulness: Learning Unconditional Uncertainty of Large Language Models](https://md.research.elamur.ai/free/en/papers/vazhentsev-2025-unconditional-truthfulness-learning-unconditional-uncertainty/)** *(2025 · 10 citations)* LLMs score each new token assuming everything generated so far is correct. If an early token is wrong, the tokens that follow can still look highly confident, so uncertainty estimates built on raw probabilities understate the risk. The authors argue that what matters is a token's unconditional chance of being… 48. **[BiMediX2 : Bio-Medical EXpert LMM for Diverse Medical Modalities](https://md.research.elamur.ai/free/en/papers/mullappilly-2025-bimedix2-bio-medical-expert-lmm/)** *(2025 · 5 citations)* BiMediX2 is a medical assistant model that works with both text and images and converses in English and Arabic. 49. **[Do Diacritics Matter? Evaluating the Impact of Arabic Diacritics on Tokenization and LLM Benchmarks](https://md.research.elamur.ai/free/en/papers/inoue-2026-do-diacritics-matter-evaluating-impact/)** *(2026 · 4 citations)* The paper asks how Arabic diacritics (vowel marks and related signs) change tokenization and the scores of large language models. 50. **[COMMUNITYNOTES: A Dataset for Exploring the Helpfulness of Fact-Checking Explanations](https://md.research.elamur.ai/free/en/papers/xing-2026-communitynotes-dataset-exploring-helpfulness-fact/)** *(2026 · 2 citations)* Platforms such as X, Meta and TikTok increasingly rely on crowd-written notes rather than professional fact-checkers to explain why a post may mislead. [← prev](https://md.research.elamur.ai/free/en/authors/timothy-baldwin/) [1](https://md.research.elamur.ai/free/en/authors/timothy-baldwin/) **2** [3](https://md.research.elamur.ai/free/en/authors/timothy-baldwin/page/3/) [next →](https://md.research.elamur.ai/free/en/authors/timothy-baldwin/page/3/)