Beneath the Surface: Investigating LLMs' Capabilities for Communicating with Subtext

• URL: http://arxiv.org/abs/2604.05273v1
• Date: Tue, 07 Apr 2026 00:14:33 GMT
• Title: Beneath the Surface: Investigating LLMs' Capabilities for Communicating with Subtext
• Abstract summary: We systematically study whether language models can use subtext in communicative settings.<n>Our evaluation settings range from writing & interpreting allegories to playing multi-agent and multi-modal games.
• Score: 14.367312995215393
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: Human communication is fundamentally creative, and often makes use of subtext -- implied meaning that goes beyond the literal content of the text. Here, we systematically study whether language models can use subtext in communicative settings, and introduce four new evaluation suites to assess these capabilities. Our evaluation settings range from writing & interpreting allegories to playing multi-agent and multi-modal games inspired by the rules of board games like Dixit. We find that frontier models generally exhibit a strong bias towards overly literal, explicit communication, and thereby fail to account for nuanced constraints -- even the best performing models generate literal clues 60% of times in one of our environments -- Visual Allusions. However, we find that some models can sometimes make use of common ground with another party to help them communicate with subtext, achieving 30%-50% reduction in overly literal clues; but they struggle at inferring presence of a common ground when not explicitly stated. For allegory understanding, we find paratextual and persona conditions to significantly shift the interpretation of subtext. Overall, our work provides quantifiable measures for an inherently complex and subjective phenomenon like subtext and reveals many weaknesses and idiosyncrasies of current LLMs. We hope this research to inspire future work towards socially grounded creative communication and reasoning.

Related papers

• Multimodal QUD: Inquisitive Questions from Scientific Figures [63.41049609329304]
  We generate inquisitive questions that reach the depth of questions humans generate when engaging with scientific papers.<n>We extend the linguistic theory of Questions Under Discussion (QUD) from being text-only to multimodal.<n>We show that fine-tuning a VLM on MQUD shifts the model from generating generic low-level visual questions to content-specific grounding.
  arXiv  Detail & Related papers  (2026-04-26T14:25:23Z)
• Can visual language models resolve textual ambiguity with visual cues? Let visual puns tell you! [14.84123301554462]
  We present UNPIE, a novel benchmark designed to assess the impact of multimodal inputs in resolving lexical ambiguities. Our dataset includes 1,000 puns, each accompanied by an image that explains both meanings. The results indicate that various Socratic Models and Visual-Language Models improve over the text-only models when given visual context.
  arXiv  Detail & Related papers  (2024-10-01T19:32:57Z)
• Is A Picture Worth A Thousand Words? Delving Into Spatial Reasoning for Vision Language Models [37.44286562901589]
  We propose SpatialEval, a novel benchmark that covers diverse aspects of spatial reasoning. We conduct a comprehensive evaluation of competitive language and vision-language models. Our findings reveal several counter-intuitive insights that have been overlooked in the literature.
  arXiv  Detail & Related papers  (2024-06-21T03:53:37Z)
• ConTextual: Evaluating Context-Sensitive Text-Rich Visual Reasoning in Large Multimodal Models [92.60282074937305]
  We introduce ConTextual, a novel dataset featuring human-crafted instructions that require context-sensitive reasoning for text-rich images. We conduct experiments to assess the performance of 14 foundation models and establish a human performance baseline. We observe a significant performance gap of 30.8% between GPT-4V and human performance.
  arXiv  Detail & Related papers  (2024-01-24T09:07:11Z)
• Natural Language Decompositions of Implicit Content Enable Better Text Representations [52.992875653864076]
  We introduce a method for the analysis of text that takes implicitly communicated content explicitly into account.<n>We use a large language model to produce sets of propositions that are inferentially related to the text that has been observed.<n>Our results suggest that modeling the meanings behind observed language, rather than the literal text alone, is a valuable direction for NLP.
  arXiv  Detail & Related papers  (2023-05-23T23:45:20Z)
• OCRBench: On the Hidden Mystery of OCR in Large Multimodal Models [122.27878464009181]
  We conducted a comprehensive evaluation of Large Multimodal Models, such as GPT4V and Gemini, in various text-related visual tasks. OCRBench contains 29 datasets, making it the most comprehensive OCR evaluation benchmark available.
  arXiv  Detail & Related papers  (2023-05-13T11:28:37Z)
• The Goldilocks of Pragmatic Understanding: Fine-Tuning Strategy Matters for Implicature Resolution by LLMs [26.118193748582197]
  We evaluate four categories of widely used state-of-the-art models. We find that, despite only evaluating on utterances that require a binary inference, models in three of these categories perform close to random. These results suggest that certain fine-tuning strategies are far better at inducing pragmatic understanding in models.
  arXiv  Detail & Related papers  (2022-10-26T19:04:23Z)
• Testing the Ability of Language Models to Interpret Figurative Language [69.59943454934799]
  Figurative and metaphorical language are commonplace in discourse. It remains an open question to what extent modern language models can interpret nonliteral phrases. We introduce Fig-QA, a Winograd-style nonliteral language understanding task.
  arXiv  Detail & Related papers  (2022-04-26T23:42:22Z)
• Emergent Communication of Generalizations [13.14792537601313]
  We argue that communicating about a single object in a shared visual context is prone to overfitting and does not encourage language useful beyond concrete reference. We propose games that require communicating generalizations over sets of objects representing abstract visual concepts. We find that these games greatly improve systematicity and interpretability of the learned languages.
  arXiv  Detail & Related papers  (2021-06-04T19:02:18Z)
• Reading and Acting while Blindfolded: The Need for Semantics in Text Game Agents [18.743819704859703]
  It remains unclear to what extent artificial agents utilize semantic understanding of the text. We propose an inverse dynamics decoder to regularize the representation space and encourage exploration. We discuss the implications of our findings for designing future agents with stronger semantic understanding.
  arXiv  Detail & Related papers  (2021-03-25T01:35:27Z)
• Interpretable Multi-Head Self-Attention model for Sarcasm Detection in social media [0.0]
  Inherent ambiguity in sarcastic expressions, make sarcasm detection very difficult. We develop an interpretable deep learning model using multi-head self-attention and gated recurrent units. We show the effectiveness of our approach by achieving state-of-the-art results on multiple datasets.
  arXiv  Detail & Related papers  (2021-01-14T21:39:35Z)
• A Benchmark for Systematic Generalization in Grounded Language Understanding [61.432407738682635]
  Humans easily interpret expressions that describe unfamiliar situations composed from familiar parts. Modern neural networks, by contrast, struggle to interpret novel compositions. We introduce a new benchmark, gSCAN, for evaluating compositional generalization in situated language understanding.
  arXiv  Detail & Related papers  (2020-03-11T08:40:15Z)
• A Deep Neural Framework for Contextual Affect Detection [51.378225388679425]
  A short and simple text carrying no emotion can represent some strong emotions when reading along with its context. We propose a Contextual Affect Detection framework which learns the inter-dependence of words in a sentence.
  arXiv  Detail & Related papers  (2020-01-28T05:03:15Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.