mlsanityData to 7 Oct 2026

Define, Evaluate, and Improve Task-Oriented Cognitive Capabilities for Instruction Generation Models

arXiv 2301.05149 ↗First posted 21 Dec 2022cs.CL cs.AI cs.HCcited by 7

Not on the map. Why?

Abstract (arXiv)

Recent work studies the cognitive capabilities of language models through psychological tests designed for humans. While these studies are helpful for understanding the general capabilities of these models, there is no guarantee that a model possessing sufficient capabilities to pass those tests would actually use those capabilities in performing real-life tasks. In this work, we formulate task-oriented cognitive capabilities, which are human-like cognitive capabilities that language models leverage to perform tasks. These capabilities are (i) the ability to quickly generate good candidate utterances (the search capability) (ii) the ability to predict how a listener interprets those utterances and choose the most appropriate one (the pragmatic capability). We design an evaluation scheme for comparing these capabilities of a language model with those of a human. Applying this scheme to examine various models in a navigation instruction generation problem, we find that their pragmatic capability is severely lacking. This insight leads us to augment them with better models of the listener and obtain a significant boost of 11% in success rate in guiding real humans. Our work advocates for having a principled procedure for aligning language models with humans that involves (i) formulating task-oriented capabilities, (ii) devising a method to quantify their deficiency, and (iii) iteratively improving them.

Key facts. Cited by 7 papers in the map (data through 7 Oct 2026). Builds on 34 of its 93 references that are in the map. First posted on arXiv 21 Dec 2022.

Builds on

34 of 93 references.

References · 34

  1. 2017Attention Is All You Needcited by 59,932
  2. 2015Deep Residual Learning for Image Recognitioncited by 51,809
  3. 2021Learning Transferable Visual Models From Natural Language Supervisioncited by 33,456
  4. 2016Simple and Scalable Predictive Uncertainty Estimation using Deep Ensemblescited by 3,888
  5. 2018CommonsenseQA: A Question Answering Challenge Targeting Commonsense Knowledgecited by 1,793
  6. 2017Vision-and-Language Navigation: Interpreting visually-grounded navigation instructions in real environmentscited by 1,013
  7. 2018From Recognition to Cognition: Visual Commonsense Reasoningcited by 553
  8. 2018On Evaluation of Embodied Navigation Agentscited by 526
  9. 2020Room-Across-Room: Multilingual Vision-and-Language Navigation with Dense Spatiotemporal Groundingcited by 401
  10. 2018Speaker-Follower Models for Vision-and-Language Navigationcited by 337
  11. 2021How Much Can CLIP Benefit Vision-and-Language Tasks?cited by 321
  12. 2020A Recurrent Vision-and-Language BERT for Navigationcited by 300
22 more
  1. 2019Learning to Navigate Unseen Environments: Back Translation with Environmental Dropoutcited by 241
  2. 2022Neural Theory-of-Mind? On the Limits of Social Intelligence in Large LMscited by 220
  3. 2020Improving Vision-and-Language Navigation with Image-Text Pairs from the Webcited by 177
  4. 2019Stay on the Path: Instruction Fidelity in Vision-and-Language Navigationcited by 129
  5. 2019General Evaluation for Instruction Conditioned Navigation using Dynamic Time Warpingcited by 102
  6. 2022A fine-grained comparison of pragmatic language understanding in humans and language modelscited by 69
  7. 2016Reasoning About Pragmatics with Neural Listeners and Speakerscited by 68
  8. 2020Multi-agent Communication meets Natural Language: Synergies between Functional and Structural Language Learningcited by 65
  9. 2018Evaluating Theory of Mind in Question Answeringcited by 64
  10. 2021Less is More: Generating Grounded Navigation Instructions from Landmarkscited by 62
  11. 2022A New Path: Scaling Vision-and-Language Navigation with Synthetic Instructions and Imitation Learningcited by 60
  12. 2017Unified Pragmatic Models for Generating and Following Instructionscited by 55
  13. 2022Inferring Rewards from Language in Contextcited by 50
  14. 2019The PhotoBook Dataset: Building Common Ground through Visually-Grounded Dialoguecited by 46
  15. 2021On the Evaluation of Vision-and-Language Navigation Instructionscited by 45
  16. 2019A Natural Language Corpus of Common Grounding under Continuous and Partially-Observable Contextcited by 35
  17. 2020RMM: A Recursive Mental Model for Dialog Navigationcited by 32
  18. 2021Few-shot Language Coordination by Modeling Theory of Mindcited by 32
  19. 2019A mathematical theory of cooperative communicationcited by 23
  20. 2022How to talk so AI will learn: Instructions, descriptions, and autonomycited by 22
  21. 2021Draw Me a Flower: Processing and Grounding Abstraction in Natural Languagecited by 14
  22. 2022Learning to Mediate Disparities Towards Pragmatic Communicationcited by 14

Built on by

Cited by year (2026 to 7 Oct)

Citing papers per year: 2023: 3; 2024: 2; 2025: 1; 2026: 1.

7 papers.

  1. 2024Position: Towards Bidirectional Human-AI Alignmentcited by 15
  2. 2023Language Models are Bounded Pragmatic Speakers: Understanding RLHF from a Bayesian Cognitive Modeling Perspectivecited by 10
  3. 2025Vision-Language Models Are Not Pragmatically Competent in Referring Expression Generationcited by 8
  4. 2023Hallucination Detection for Grounded Instruction Generationcited by 7
  5. 2024Successfully Guiding Humans with Imperfect Instructions by Highlighting Potential Errors and Suggesting Correctionscited by 2
  6. 2023Utility-based Adaptive Teaching Strategies using Bayesian Theory of Mindcited by 1
  7. 2026Pragmatics Meets Culture: Culturally-adapted Artwork Description Generation and Evaluationcited by 0

Similar, not cited

Close in wording, no citation either way

  1. 2024Auxiliary task demands mask the capabilities of smaller language models
  2. 2024From Babbling to Fluency: Evaluating the Evolution of Language Models in Terms of Human Language Acquisition
  3. 2023Psychometric Predictive Power of Large Language Models
  4. 2023Reasoning or Reciting? Exploring the Capabilities and Limitations of Language Models Through Counterfactual Tasks
  5. 2025Can Language Models Compose Skills In-Context?
  6. 2026Language Model Goal Selection Differs from Humans' in a Self-Directed Learning Task
  7. 2023Did You Read the Instructions? Rethinking the Effectiveness of Task Definitions in Instruction Learning
  8. 2024Hierarchical Prompting Taxonomy: A Universal Evaluation Framework for Large Language Models Aligned with Human Cognitive Principles