Proceedings of the Seventh Workshop on Insights from Negative Results in NLP

Proceedings of the Seventh Workshop on Insights from Negative Results in NLP João Sedoc editor Raphael Shu editor Aleksandr Drozd editor Arjun Akula editor Anna Rogers editor Salvatore Giorgi editor Adrian Benton editor 2026-10 Association for Computational Linguistics Budapest, Hungary text conference publication 979-8-89176-480-4 insights-2026-1 https://aclanthology.org/2026.insights-1.0/ Geometric Separation Without Functional Specialization: Auditing Subspace Disentanglement in Dual-Encoder LLM Paraphrase Detection Govind Arun author Ashish Abraham author Gadha Lekshmi P author 2026-10 text Proceedings of the Seventh Workshop on Insights from Negative Results in NLP João Sedoc editor Raphael Shu editor Aleksandr Drozd editor Arjun Akula editor Anna Rogers editor Salvatore Giorgi editor Adrian Benton editor Association for Computational Linguistics Budapest, Hungary conference publication 979-8-89176-480-4 arun-etal-2026-geometric https://aclanthology.org/2026.insights-1.1/ 2026-10 1 6 Exploring Dowker Homology for Sentence Similarity Marius Huber author Juri Opitz author 2026-10 text Proceedings of the Seventh Workshop on Insights from Negative Results in NLP João Sedoc editor Raphael Shu editor Aleksandr Drozd editor Arjun Akula editor Anna Rogers editor Salvatore Giorgi editor Adrian Benton editor Association for Computational Linguistics Budapest, Hungary conference publication 979-8-89176-480-4 huber-opitz-2026-exploring https://aclanthology.org/2026.insights-1.2/ 2026-10 7 18 Does Pragmatic Transfer Help Euphemism Detection? Evidence from Target-Span Marking Whitney Poh author Noriko Takahashi author Patrick Lee author Julia Sammartino author Libby Barak author Jing Peng author Anna Feldman author 2026-10 text Proceedings of the Seventh Workshop on Insights from Negative Results in NLP João Sedoc editor Raphael Shu editor Aleksandr Drozd editor Arjun Akula editor Anna Rogers editor Salvatore Giorgi editor Adrian Benton editor Association for Computational Linguistics Budapest, Hungary conference publication 979-8-89176-480-4 poh-etal-2026-pragmatic https://aclanthology.org/2026.insights-1.3/ 2026-10 19 32 Risk-Averse Failure Modes of Jensen-Shannon Distance in RLVR Distributional Alignment Sebastian Loftus author Xinpeng Wang author Marie-Catherine De Marneffe author Barbara Plank author 2026-10 text Proceedings of the Seventh Workshop on Insights from Negative Results in NLP João Sedoc editor Raphael Shu editor Aleksandr Drozd editor Arjun Akula editor Anna Rogers editor Salvatore Giorgi editor Adrian Benton editor Association for Computational Linguistics Budapest, Hungary conference publication 979-8-89176-480-4 loftus-etal-2026-risk https://aclanthology.org/2026.insights-1.4/ 2026-10 33 42 Controlled Fine-Tuning for Two-Tower Recommenders: Why Gradual Unfreezing Can Fail and Cosine Regularization Suffices Gabriel Dershowitz author Djallel Bouneffouf author 2026-10 text Proceedings of the Seventh Workshop on Insights from Negative Results in NLP João Sedoc editor Raphael Shu editor Aleksandr Drozd editor Arjun Akula editor Anna Rogers editor Salvatore Giorgi editor Adrian Benton editor Association for Computational Linguistics Budapest, Hungary conference publication 979-8-89176-480-4 dershowitz-bouneffouf-2026-controlled https://aclanthology.org/2026.insights-1.5/ 2026-10 43 47 Do Soft Prompts Have Useful Wavelet Structure? Negative Evidence In A Controlled Setup Soumyadeep Dey author Srikanth Tenneti author Bamdev Mishra author Shobhit Niranjan author Mayur Datar author 2026-10 text Proceedings of the Seventh Workshop on Insights from Negative Results in NLP João Sedoc editor Raphael Shu editor Aleksandr Drozd editor Arjun Akula editor Anna Rogers editor Salvatore Giorgi editor Adrian Benton editor Association for Computational Linguistics Budapest, Hungary conference publication 979-8-89176-480-4 dey-etal-2026-soft https://aclanthology.org/2026.insights-1.6/ 2026-10 48 59 Decomposing LLM-Judge–Human Agreement: How Much Is Inherited Surface Preference? Gaurav Kumar author 2026-10 text Proceedings of the Seventh Workshop on Insights from Negative Results in NLP João Sedoc editor Raphael Shu editor Aleksandr Drozd editor Arjun Akula editor Anna Rogers editor Salvatore Giorgi editor Adrian Benton editor Association for Computational Linguistics Budapest, Hungary conference publication 979-8-89176-480-4 kumar-2026-decomposing https://aclanthology.org/2026.insights-1.7/ 2026-10 60 69 The Flip-Test: A Behavioral Diagnostic for Multi-Agent Debate Failures Julia Hu author Alfred Shen author Kumar Lakshmipathi author 2026-10 text Proceedings of the Seventh Workshop on Insights from Negative Results in NLP João Sedoc editor Raphael Shu editor Aleksandr Drozd editor Arjun Akula editor Anna Rogers editor Salvatore Giorgi editor Adrian Benton editor Association for Computational Linguistics Budapest, Hungary conference publication 979-8-89176-480-4 hu-etal-2026-flip https://aclanthology.org/2026.insights-1.8/ 2026-10 70 75 Activation-Based Active Learning for In-Context Learning: Challenges and Insights Yaseen Osman author Geoff Merrett author Stuart Middleton author 2026-10 text Proceedings of the Seventh Workshop on Insights from Negative Results in NLP João Sedoc editor Raphael Shu editor Aleksandr Drozd editor Arjun Akula editor Anna Rogers editor Salvatore Giorgi editor Adrian Benton editor Association for Computational Linguistics Budapest, Hungary conference publication 979-8-89176-480-4 osman-etal-2026-activation https://aclanthology.org/2026.insights-1.9/ 2026-10 76 83 The Architecture Was Doing the Work: Why NLI Debiasing Results Do Not Generalise Prabhjot Singh author 2026-10 text Proceedings of the Seventh Workshop on Insights from Negative Results in NLP João Sedoc editor Raphael Shu editor Aleksandr Drozd editor Arjun Akula editor Anna Rogers editor Salvatore Giorgi editor Adrian Benton editor Association for Computational Linguistics Budapest, Hungary conference publication 979-8-89176-480-4 singh-2026-architecture https://aclanthology.org/2026.insights-1.10/ 2026-10 84 89 When Learned Context Planning Fails to Beat Strong Retrieval: A Controlled Study of Planning, Routing, and Reranking for Long-Context QA Yingrui Li author Han Chen author 2026-10 text Proceedings of the Seventh Workshop on Insights from Negative Results in NLP João Sedoc editor Raphael Shu editor Aleksandr Drozd editor Arjun Akula editor Anna Rogers editor Salvatore Giorgi editor Adrian Benton editor Association for Computational Linguistics Budapest, Hungary conference publication 979-8-89176-480-4 li-chen-2026-learned https://aclanthology.org/2026.insights-1.11/ 2026-10 90 94 No Optimal Language Set Exists for Multilingual Instruction Tuning: Insights from a Linguistically-Informed Study Gürkan Soykan author Gözde Şahin author 2026-10 text Proceedings of the Seventh Workshop on Insights from Negative Results in NLP João Sedoc editor Raphael Shu editor Aleksandr Drozd editor Arjun Akula editor Anna Rogers editor Salvatore Giorgi editor Adrian Benton editor Association for Computational Linguistics Budapest, Hungary conference publication 979-8-89176-480-4 soykan-sahin-2026-optimal https://aclanthology.org/2026.insights-1.12/ 2026-10 95 101 Short-Context Gains, Long-Context Drift: Negative Results for Token-Conditioned RoPE Allan Kazakov author 2026-10 text Proceedings of the Seventh Workshop on Insights from Negative Results in NLP João Sedoc editor Raphael Shu editor Aleksandr Drozd editor Arjun Akula editor Anna Rogers editor Salvatore Giorgi editor Adrian Benton editor Association for Computational Linguistics Budapest, Hungary conference publication 979-8-89176-480-4 kazakov-2026-short https://aclanthology.org/2026.insights-1.13/ 2026-10 102 109 When Gradient Importance Lies: Adaptive LoRA Rank Allocation Fails Under GRPO Yash Sawant author 2026-10 text Proceedings of the Seventh Workshop on Insights from Negative Results in NLP João Sedoc editor Raphael Shu editor Aleksandr Drozd editor Arjun Akula editor Anna Rogers editor Salvatore Giorgi editor Adrian Benton editor Association for Computational Linguistics Budapest, Hungary conference publication 979-8-89176-480-4 sawant-2026-gradient https://aclanthology.org/2026.insights-1.14/ 2026-10 110 114 Do Not Score the Edited Answer: A Negative Result on Loose Evaluation for Verifiable Instruction Following José Luis Delgado author 2026-10 text Proceedings of the Seventh Workshop on Insights from Negative Results in NLP João Sedoc editor Raphael Shu editor Aleksandr Drozd editor Arjun Akula editor Anna Rogers editor Salvatore Giorgi editor Adrian Benton editor Association for Computational Linguistics Budapest, Hungary conference publication 979-8-89176-480-4 delgado-2026-score https://aclanthology.org/2026.insights-1.15/ 2026-10 115 119 Perturbation, Not Provenance: Why Activation Steering Does Not Make a Robust Text Watermark Sergey Pletenev author 2026-10 text Proceedings of the Seventh Workshop on Insights from Negative Results in NLP João Sedoc editor Raphael Shu editor Aleksandr Drozd editor Arjun Akula editor Anna Rogers editor Salvatore Giorgi editor Adrian Benton editor Association for Computational Linguistics Budapest, Hungary conference publication 979-8-89176-480-4 pletenev-2026-perturbation https://aclanthology.org/2026.insights-1.16/ 2026-10 120 126 PEFT Gains Need Hidden Baselines: Rank and Optimizer Sweeps Change LoRA Conclusions Ding Wu author 2026-10 text Proceedings of the Seventh Workshop on Insights from Negative Results in NLP João Sedoc editor Raphael Shu editor Aleksandr Drozd editor Arjun Akula editor Anna Rogers editor Salvatore Giorgi editor Adrian Benton editor Association for Computational Linguistics Budapest, Hungary conference publication 979-8-89176-480-4 wu-2026-peft https://aclanthology.org/2026.insights-1.17/ 2026-10 127 132 When Relation-Free Retrieval Is Not Relation-Aware: A Directionality Stress Test for Linear Graph RAG Ding Wu author 2026-10 text Proceedings of the Seventh Workshop on Insights from Negative Results in NLP João Sedoc editor Raphael Shu editor Aleksandr Drozd editor Arjun Akula editor Anna Rogers editor Salvatore Giorgi editor Adrian Benton editor Association for Computational Linguistics Budapest, Hungary conference publication 979-8-89176-480-4 wu-2026-relation https://aclanthology.org/2026.insights-1.18/ 2026-10 133 136