The Gold Medals in an Empty Room: Diagnosing Metalinguistic Reasoning in LLMs with Camlang

• URL: http://arxiv.org/abs/2509.00425v1
• Date: Sat, 30 Aug 2025 09:13:10 GMT
• Title: The Gold Medals in an Empty Room: Diagnosing Metalinguistic Reasoning in LLMs with Camlang
• Abstract summary: Large Language Models (LLMs) achieve gold-medal performance across many benchmarks, yet it remains unclear whether such success reflects genuine reasoning or pattern matching.<n>From a cognitive science perspective, an informative test is whether models can master an unfamiliar language through explicit metalinguistic deductive learning.<n>We address this question with Camlang, a novel constructed language that exhibits naturalistic yet unattested feature combinations.
• Score: 8.440204591615442
• License: http://creativecommons.org/licenses/by-nc-sa/4.0/
• Abstract: Large Language Models (LLMs) achieve gold-medal performance across many benchmarks, yet it remains unclear whether such success reflects genuine reasoning or pattern matching. From a cognitive science perspective, an informative test is whether models can master an unfamiliar language through explicit metalinguistic deductive learning, a paradigm where human learners can reliably internalise grammatical systems through metalinguistic reasoning. We address this question with Camlang, a novel constructed language that exhibits naturalistic yet unattested feature combinations. Camlang consists of two explicit resources, a grammar book and a bilingual dictionary, which mirror adult second-language learning via explicit grammar rules and lexical lookup, and enable us to disentangle errors in morpho-syntax, lexical semantics, and sentence-level reasoning. Human experiments show that these resources are sufficient for participants to acquire Camlang and successfully solve Camlang tasks. To operationalise evaluation, we adapt CommonsenseQA into Camlang, creating Camlang-CSQA-v0, the first task in a broader suite where solving questions requires applying grammar rules and lexical mappings. Experimental results show that GPT-5 achieves 98\% EM accuracy in English but only 47\% in Camlang, far below human performance at 87\%, while other state-of-the-art reasoning LLMs perform even worse. Human verification further reveals that most model successes stem from shallow lexical alignment while GPT-5 shows emerging metalinguistic awareness to a limited extent but not systematic grammatical mastery as humans. Camlang establishes a cognitively grounded evaluation paradigm that exposes fundamental gaps between current models and human metalinguistic competence.

Related papers

• LLM Reasoning Predicts When Models Are Right: Evidence from Coding Classroom Discourse [0.18268488712787334]
  Large Language Models (LLMs) are increasingly deployed to automatically label and analyze educational dialogue at scale.<n>We investigate whether reasoning generated by LLMs can be used to predict the correctness of a model's own predictions.<n>We analyze 30,300 teacher utterances from classroom dialogue, each labeled by multiple state-of-the-art LLMs with an instructional move construct and an accompanying reasoning.
  arXiv  Detail & Related papers  (2026-02-10T14:38:13Z)
• Large Language Models as Neurolinguistic Subjects: Discrepancy between Performance and Competence [49.60849499134362]
  This study investigates the linguistic understanding of Large Language Models (LLMs) regarding signifier (form) and signified (meaning)<n>We introduce a neurolinguistic approach, utilizing a novel method that combines minimal pair and diagnostic probing to analyze activation patterns across model layers.<n>We found: (1) Psycholinguistic and neurolinguistic methods reveal that language performance and competence are distinct; (2) Direct probability measurement may not accurately assess linguistic competence; and (3) Instruction tuning won't change much competence but improve performance.
  arXiv  Detail & Related papers  (2024-11-12T04:16:44Z)
• Holmes: A Benchmark to Assess the Linguistic Competence of Language Models [59.627729608055006]
  We introduce Holmes, a new benchmark designed to assess language models (LMs) linguistic competence. We use computation-based probing to examine LMs' internal representations regarding distinct linguistic phenomena. As a result, we meet recent calls to disentangle LMs' linguistic competence from other cognitive abilities.
  arXiv  Detail & Related papers  (2024-04-29T17:58:36Z)
• PhonologyBench: Evaluating Phonological Skills of Large Language Models [57.80997670335227]
  Phonology, the study of speech's structure and pronunciation rules, is a critical yet often overlooked component in Large Language Model (LLM) research. We present PhonologyBench, a novel benchmark consisting of three diagnostic tasks designed to explicitly test the phonological skills of LLMs. We observe a significant gap of 17% and 45% on Rhyme Word Generation and Syllable counting, respectively, when compared to humans.
  arXiv  Detail & Related papers  (2024-04-03T04:53:14Z)
• A Sentence is Worth a Thousand Pictures: Can Large Language Models Understand Hum4n L4ngu4ge and the W0rld behind W0rds? [2.7342737448775534]
  Large Language Models (LLMs) have been linked to claims about human-like linguistic performance. We analyze the contribution of LLMs as theoretically informative representations of a target cognitive system. We evaluate the models' ability to see the bigger picture, through top-down feedback from higher levels of processing.
  arXiv  Detail & Related papers  (2023-07-26T18:58:53Z)
• Simple Linguistic Inferences of Large Language Models (LLMs): Blind Spots and Blinds [59.71218039095155]
  We evaluate language understanding capacities on simple inference tasks that most humans find trivial. We target (i) grammatically-specified entailments, (ii) premises with evidential adverbs of uncertainty, and (iii) monotonicity entailments. The models exhibit moderate to low performance on these evaluation sets.
  arXiv  Detail & Related papers  (2023-05-24T06:41:09Z)
• ChatABL: Abductive Learning via Natural Language Interaction with ChatGPT [72.83383437501577]
  Large language models (LLMs) have recently demonstrated significant potential in mathematical abilities. LLMs currently have difficulty in bridging perception, language understanding and reasoning capabilities. This paper presents a novel method for integrating LLMs into the abductive learning framework.
  arXiv  Detail & Related papers  (2023-04-21T16:23:47Z)
• Interpretable Unified Language Checking [42.816372695828306]
  We present an interpretable, unified, language checking (UniLC) method for both human and machine-generated language. We find that LLMs can achieve high performance on a combination of fact-checking, stereotype detection, and hate speech detection tasks.
  arXiv  Detail & Related papers  (2023-04-07T16:47:49Z)
• Testing AI on language comprehension tasks reveals insensitivity to underlying meaning [3.335047764053173]
  Large Language Models (LLMs) are recruited in applications that span from clinical assistance and legal support to question answering and education. Yet, reverse-engineering is bound by Moravec's Paradox, according to which easy skills are hard. We systematically assess 7 state-of-the-art models on a novel benchmark.
  arXiv  Detail & Related papers  (2023-02-23T20:18:52Z)
• Towards Language Modelling in the Speech Domain Using Sub-word Linguistic Units [56.52704348773307]
  We propose a novel LSTM-based generative speech LM based on linguistic units including syllables and phonemes. With a limited dataset, orders of magnitude smaller than that required by contemporary generative models, our model closely approximates babbling speech. We show the effect of training with auxiliary text LMs, multitask learning objectives, and auxiliary articulatory features.
  arXiv  Detail & Related papers  (2021-10-31T22:48:30Z)
• Masked Language Modeling and the Distributional Hypothesis: Order Word Matters Pre-training for Little [74.49773960145681]
  A possible explanation for the impressive performance of masked language model (MLM)-training is that such models have learned to represent the syntactic structures prevalent in NLP pipelines. In this paper, we propose a different explanation: pre-trains succeed on downstream tasks almost entirely due to their ability to model higher-order word co-occurrence statistics. Our results show that purely distributional information largely explains the success of pre-training, and underscore the importance of curating challenging evaluation datasets that require deeper linguistic knowledge.
  arXiv  Detail & Related papers  (2021-04-14T06:30:36Z)

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.