--- title: "Yonatan Belinkov — papers, page 2 · Harvard University" description: "Yonatan Belinkov, Harvard University: 37 papers on Elamur Research (2014–2026), h-index 43 — affiliations, topics, co-authors and who…" canonical: "https://research.elamur.ai/free/en/authors/yonatan-belinkov/page/2/" --- [Elamur Research](https://md.research.elamur.ai/free/en/) · [this page in HTML](https://research.elamur.ai/free/en/authors/yonatan-belinkov/page/2/) [home](https://md.research.elamur.ai/free/en/) / [authors](https://md.research.elamur.ai/free/en/authors/) / [Yonatan Belinkov](https://md.research.elamur.ai/free/en/authors/yonatan-belinkov/) / page 2 # Yonatan Belinkov · page 2 37 papers in our library · 2014–2026 · page 2 of 2 [The profile, affiliations and co-authors →](https://md.research.elamur.ai/free/en/authors/yonatan-belinkov/) ## Papers The most cited first (OpenAlex and our library). [← prev](https://md.research.elamur.ai/free/en/authors/yonatan-belinkov/) [1](https://md.research.elamur.ai/free/en/authors/yonatan-belinkov/) **2** 26. **[Back Attention: Understanding and Enhancing Multi-Hop Reasoning in Large Language Models](https://md.research.elamur.ai/free/en/papers/yu-2025-back-attention-understanding-enhancing-multi/)** *(2025 · 16 citations)* The paper examines why large language models often fail at latent multi-hop factual reasoning, where a query such as "e1's r1's r2 is" has to be answered without explicit intermediate steps. 27. **[Measuring Chain of Thought Faithfulness by Unlearning Reasoning Steps](https://md.research.elamur.ai/free/en/papers/tutek-2025-measuring-chain-thought-faithfulness-unlearning/)** *(2025 · 15 citations)* The paper asks whether the step-by-step reasoning (chain of thought, CoT) that language models produce reflects the knowledge the model actually relies on. 28. **[Large-Scale Machine Translation between Arabic and Hebrew: Available Corpora and Initial Results](https://md.research.elamur.ai/free/en/papers/belinkov-2016-large-scale-machine-translation-between/)** *(2016 · 14 citations)* The authors test Arabic-to-Hebrew machine translation on the WIT3 TED talks corpus (about 0.2M training sentences, 3.2M Arabic words), with the IWSLT 2016 tuning and test splits. 29. **[SAEs Are Good for Steering – If You Select the Right Features](https://md.research.elamur.ai/free/en/papers/arad-2025-saes-are-good-steering-if/)** *(2025 · 12 citations)* Sparse autoencoders (SAEs) break a language model's hidden states into interpretable features. Those features can be amplified to steer generation toward a concept without labeled data. Earlier evaluations such as AxBench, however, found SAE steering weak. The authors argue that the problem lies in how features are… 30. **[Studying the History of the Arabic Language: Language Technology and a Large-Scale Historical Corpus](https://md.research.elamur.ai/free/en/papers/belinkov-2018-studying-history-arabic-language-language/)** *(2018 · 12 citations)* Written Arabic was standardized very early and is often described as almost unchanged for over a millennium. 31. **[A Generative Approach for Mitigating Structural Biases in Natural Language Inference](https://md.research.elamur.ai/free/en/papers/asael-2022-generative-approach-mitigating-structural-biases/)** *(2022 · 10 citations)* NLI benchmarks such as SNLI and MNLI allow classifiers to score well while using only part of the input, such as the hypothesis by itself or the words shared by premise and hypothesis. 32. **[Are Formal and Functional Linguistic Mechanisms Dissociated in Language Models?](https://md.research.elamur.ai/free/en/papers/hanna-2026-are-formal-functional-linguistic-mechanisms/)** *(2026 · 5 citations)* This study asks whether today's LLMs internally separate formal linguistic competence (syntax, morphology, lexical semantics) from functional competence (reasoning, world knowledge, tracking situations). 33. **[Answer Selection in Arabic Community Question Answering: A Feature-Rich Approach](https://md.research.elamur.ai/free/en/papers/belinkov-2015-answer-selection-arabic-community-question/)** *(2015 · 5 citations)* The paper tackles answer selection in Arabic community question answering. Given a forum question and its attached comments, each comment must be labelled DIRECT, RELATED or IRRELEVANT. The authors use the SemEval-2015 Task 3 Arabic setup, built on the Fatwa corpus of religious questions answered by scholars. Every… 34. **[A Dataset for Metaphor Detection in Early Medieval Hebrew Poetry](https://md.research.elamur.ai/free/en/papers/toker-2024-dataset-metaphor-detection-early-medieval/)** *(2024 · 3 citations)* The authors release the first metaphor-annotated corpus of Hebrew piyyut, the liturgical poetry of the 5th–8th centuries CE, taken from Cairo Genizah reconstructions. 35. **[Mechanisms of Prompt-Induced Hallucination in Vision–Language Models](https://md.research.elamur.ai/free/en/papers/rudman-2026-mechanisms-prompt-induced-hallucination-vision/)** *(2026 · 2 citations)* Vision–language models sometimes trust a user's wording more than what the image shows. This paper studies that failure, called prompt-induced hallucination (PIH), using object counting. Images from CountBench are first counted correctly by the model. The same images are then paired with a request to describe more… 36. **[Will it Merge? On The Causes of Model Mergeability](https://md.research.elamur.ai/free/en/papers/rahamim-2026-will-it-merge-causes-model/)** *(2026)* Merging combines several fine-tuned models into one multitask model without retraining, yet it is unclear why some fine-tuned updates survive merging better than others. 37. **[Part-of-Speech and Morphological Tagging of Algerian Judeo-Arabic](https://md.research.elamur.ai/free/en/papers/becker-2022-part-speech-morphological-tagging-algerian/)** *(2022)* Algerian Judeo-Arabic (AJA) is the Arabic of Algeria's Jewish communities, written in Hebrew script. It now has very few speakers, so linguists study it mainly through printed and handwritten texts from the 19th–20th centuries. Manual morphological annotation of these texts is slow and needs rare expertise, which… [← prev](https://md.research.elamur.ai/free/en/authors/yonatan-belinkov/) [1](https://md.research.elamur.ai/free/en/authors/yonatan-belinkov/) **2**