analiza aiqbench
This research addresses the "self-supervision trap" in time-series analysis by decoupling statistical perception from linguistic description, allowing deterministic code to handle complex multivariate patterns while LLMs focus on expression. The resulting CGTime model significantly outperforms larger general-purpose models like GPT-4o-mini and GPT-5.4-nano on verifiable fact-scoring benchmarks for real-world data streams. By offloading heavy perceptual lifting to computation, this approach offers a scalable path toward reliable AI interpretation of multivariate temporal signals without relying solely on imperfect human labels or model overfitting.
Decoupling Perception from Description: Computation-Grounded Representation Alignment between Multivariate Time Series and Language
arXiv:2608.05238v1 Announce Type: new Abstract: 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 several variables, right where the labeling LLM's limits are most exposed. These two problems create a trilemma: existing methods are reliable, realistic, or scalable, but none achieves all three. We resolve this by decoupling perception from description. Deterministic code computes a set of statistics from real, open-source multivariate series; the LLM verbalizes those precomputed facts. Perception, which LLMs do poorly, is handled by computation, while the LLM handles expression. This produces CGTime, our 4B-parameter computation-grounded time-series-language model. CGTime outperforms far larger general-purpose models on multivariate understanding tasks: it attains the best multivariate fact score on our held-out benchmark (0.283 vs. 0.173 for GPT-4o-mini and 0.203 for GPT-5.4-nano), a gap that survives Holm-corrected paired significance tests against every baseline. It also states verifiable numerical facts in generated captions more accurately and covers a broader range of statistical properties.
Abstract: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 several variables, right where the labeling LLM's limits are most exposed. These two problems create a trilemma: existing methods are reliable, realistic, or scalable, but none achieves all three. We resolve this by decoupling perception from description. Deterministic code computes a set of statistics from real, open-source multivariate series; the LLM verbalizes those precomputed facts. Perception, which LLMs do poorly, is handled by computation, while the LLM handles expression. This produces CGTime, our 4B-parameter computation-grounded time-series-language model. CGTime outperforms far larger general-purpose models on multivariate understanding tasks: it attains the best multivariate fact score on our held-out benchmark (0.283 vs. 0.173 for GPT-4o-mini and 0.203 for GPT-5.4-nano), a gap that survives Holm-corrected paired significance tests against every baseline. It also states verifiable numerical facts in generated captions more accurately and covers a broader range of statistical properties.
46 pages, 9 figures, including supplementary material. Submitted to AAAI 2027. Xinran Feng and Yi Xie contributed equally
Machine Learning (cs.LG)
arXiv:2608.05238 [cs.LG]
(or arXiv:2608.05238v1 [cs.LG] for this version)
https://doi.org/10.48550/arXiv.2608.05238 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.
Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.
Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.
Text integral sub licența deschisă a publicației · 329 cuvinte
Subiecte conexe
Why do models task game?
AI Alignment Forum · 6 aug. 2026
Skip the Talk, Re-Focus on Vision: Latent Reasoning for Reasoning Segmentation in Multimodal Large Language Models
arXiv · 28 sept. 2026
Illusionists should try to build hedonium
LessWrong · 17 iun. 2026
Universal Pose Pretraining for Generalizable Vision-Language-Action Policies
arXiv · 30 sept. 2026