arXiv：cs.LG（机器学习，全量分类）· Daniel Quigley, Eric Maynard·· 17 小时前AI 评分29
一个审计 grounding 主张的框架
A framework for auditing grounding claims
AI 导读
研究者提出一个对照声明语义标准来审计 grounding 主张的框架，报告测量结果与证据，最终判定取决于明确的接受标准。在玩具 gridworld 中，智能体能准确解释单个符号，却在一个被保留的组合上失败，而按声明规则组合其解释本应成功，由此定位出对组合规则的偏离。该审计与一项预训练词向量试点均表明，指定机制对当前性能有贡献，但其是否解释该机制的保留仍未获认证。
正文
Abstract:The symbol grounding problem asks how a token such as cat can be about cats. We propose a framework for auditing grounding claims against a declared semantic standard. The audit reports measurements and evidence, with overall verdicts conditional on explicit acceptance criteria. Its profiles assess accuracy, robustness, and composition alongside evidence about how the system acquired its mechanisms, how they contribute to performance, and why they were retained. In a toy gridworld, an agent interprets individual symbols accurately but fails a withheld combination. Composing its interpretations by the declared rule would succeed. This comparison identifies a departure from the composition rule within the observed failure. Both this audit and a pilot on pretrained word vectors provide evidence that a designated mechanism contributes to present performance. Whether that contribution explains its retention remains uncertified. The framework evaluates the evidence for grounding claims; candidate accounts remain responsible for explaining how meaning emerges.
| Comments: | resubmission: 38 pages, 90 sources, 3 figures |
| Subjects: | Artificial Intelligence (cs.AI); Computation and Language (cs.CL); Machine Learning (cs.LG) |
| MSC classes: | 68T01 (Primary) 68T05, 68T30, 68T50 (Secondary) |
| ACM classes: | I.2.0; I.2.4; I.2.6; I.2.7 |
| Cite as: | arXiv:2512.06205 [cs.AI] |
| (or arXiv:2512.06205v3 [cs.AI] for this version) | |
| https://doi.org/10.48550/arXiv.2512.06205 arXiv-issued DOI via DataCite |
Submission history
From: Daniel Quigley [view email]
[v1]
Fri, 5 Dec 2025 22:58:47 UTC (74 KB)
[v2]
Wed, 31 Dec 2025 02:06:22 UTC (80 KB)
[v3]
Thu, 1 Oct 2026 01:48:05 UTC (89 KB)
来源：arXiv：cs.LG（机器学习，全量分类） · arxiv.org