{
  "id": 17397,
  "url": "https://arxiv.org/abs/2608.05238v1",
  "title": "Decoupling Perception from Description: Computation-Grounded Representation Alignment between Multivariate Time Series and Language",
  "summary": "Training multimodal models to align time series with language runs into a self-supervision trap. The usual recipe asks an LLM to read a series and write a description, so label quality is capped by the perceptual skill the model is supposed to learn. The data can never teach more than the labeler already knows. A second gap makes this worse: most datasets use a single variable, but the patterns that matter (cross-channel correlation, lead-lag structure, co-occurring anomalies) appear only with s",
  "authors": "Xinran Feng, Yi Xie, Chao Zhang, Ruikun Li, Wanyun Ling, Ziyue Li et al.",
  "category": "research",
  "topics": "safety-alignment",
  "orgs": null,
  "regions": null,
  "published_at": "2026-08-05T13:57:08.000Z",
  "fetched_at": "2026-08-07T05:10:58.501Z",
  "source_slug": "arxiv-cslg",
  "source_name": "arXiv cs.LG",
  "source_homepage": "https://arxiv.org/list/cs.LG/recent",
  "ethics_ai_record_url": "https://ethics.ai/record/17397",
  "original_url": "https://arxiv.org/abs/2608.05238v1",
  "evidence_status": "source-only",
  "attribution": "via ethics.ai"
}