AI
THPL: A Vision-to-Language Decision Support Framework for Rainbow Trout Feeding Management in RAS
Publication: cs.AI updates on arXiv.org
Author: Meng Liang, Guanbo Feng, Haozhuang Chi, Shilong Zhao, Zhixin Xiong, Yuhang He, Wenfeng Han, Tianhao Zhao, Zhihong Ma, Ying Liu
Source published: 05/10/2026, 04:00:00 UTC
Briefing created: 05/10/2026, 04:08:12 UTC
Based on a feed excerpt; full article content has not been acquired. Source evidence retrieved 05/10/2026, 04:03:25 UTC.
Briefings
This study introduces THPL, a generative framework for rainbow trout feeding management in Recirculating Aquaculture Systems (RAS). The system uses Fishsort to extract feeding trajectories, which are processed by a Hierarchical Behavior Encoder to create dual-evidence representations combining physical and implicit behavioral data. These tokens are integrated with environmental parameters and expert rules to fine-tune an LLM via LoRA and counterfactual mDPO. Results show a strong correlation between the Activity Coefficient and expert-annotated feeding intensity, with decision accuracy improving from 33.33% to 93.33% using dual-evidence tokens compared to text-only baselines.
Key points
- The THPL framework uses Fishsort to extract feeding trajectories from RAS systems to establish an Activity Coefficient (AC) quantifying feeding intensity.
- Decision accuracy improved from 33.33% to 93.33% when continuous spatiotemporal tokens were used instead of text-only baselines.
Context and limitations
The study relies on preprint data from arXiv and does not include experimental validation or field trial results.
Generated by Bittrees-hosted qwen35-2b sha256:aaf42c8b7c3c on 05/10/2026, 04:08:12 UTC. Based on publisher-provided feed evidence; may not cover the full original article. Check the source for details.
View archived briefing on IPFS
bafybeib3wdvl3dq42dyqfs2aiwg3y5bylh4j2lueixoyfsdf6ivynvfola