Graham Neubig · page 4
Papers
- An Incomplete Loop: Deductive, Inductive, and Abductive Reasoning in Language Models (2024 · 12 citations)
Language models can acquire a new task in three ways: from an explicit instruction, from a few examples, or by first writing their own task description from examples and then following it. - Grounding Multilingual Multimodal LLMs With Cultural Knowledge (2025 · 11 citations)
Multimodal LLMs tend to fail on culturally specific, long-tail entities and perform worse in low-resource languages, mainly because their training data is dominated by English and Western content. - Coding Agents with Multimodal Browsing are Generalist Problem Solvers (2026 · 5 citations)
Strong LLM agents for software engineering, deep research and web navigation are usually specialists. Each is built around tools and workflows tuned for a narrow set of benchmarks, so it breaks down outside its home domain. The authors ask whether a single agent with a small set of general-purpose tools can match… - Gained in Translation: Privileged Pairwise Judges Enhance Multilingual Reasoning (2026 · 2 citations)
Reasoning LLMs trained mostly on English chains of thought perform much worse when the same questions are posed in less-represented languages. - MERLIN: A Testbed for Multilingual Multimodal Entity Recognition and Linking (2026 · 2 citations)
Entity linking maps an ambiguous mention to its referent in a knowledge base. It becomes difficult when the surrounding text is short, as in news headlines. Until now, multilingual linking research has used text alone. This paper asks whether accompanying images can resolve such ambiguity across languages. The… - CAIRE: Cultural Attribution of Images with Retrieval (2026 · 1 citation)
Text-to-image systems often depict cultures unevenly, favouring Western imagery or producing stereotyped and inaccurate pictures.
This page in Markdown → https://md.research.elamur.ai/free/en/authors/graham-neubig/page/4/