Tag: #zero-shot10 results
A domestically developed multimodal Agent achieves state-of-the-art performance in medical image segmentation without model modifications or extra tokens. The research has been accepted to CVPR 2026, highlighting efficient agent-based approaches.
Sudo Tech, valued at $2B, launched its first embodied AI model Sudo R1. It achieves 98% first-grasp success rate in zero-shot settings using zero real machine data. This debut showcases advanced capabilities in robotic manipulation.
CourtGuard is a retrieval-augmented multi-agent framework that treats LLM safety as an evidentiary debate using external policy documents. It achieves state-of-the-art results on 7 safety benchmarks without fine-tuning, outperforming policy-following baselines. It excels in zero-shot adaptability (90% accuracy on Wikipedia Vandalism) and automated curation of 9 adversarial datasets.
Zero-shot World Model (ZWM) trained solely on one child's visual data matches state-of-the-art models zero-shot on visual-cognitive tasks. It narrows the massive data gap between AI and human children. Provides a blueprint for data-efficient AI from human-scale inputs.
Structured 5-layer prompts reduce vocabulary contamination from 80% to 5% for Tulu (low-resource Dravidian lang) without fine-tuning. Layers include phonological grounding, morphology, negative constraints, romanization, self-play examples. Validated on GPT-4o, Gemini 2.0 Flash, Llama 3.1 70B.
The new DM0.5 model from Yuanli Lingji has been released, trained on 150,000 hours of data. It demonstrates a 31% improvement in zero-shot performance and shows signs of emergent generalization capabilities.
The 'Query Retrieve Conclude' framework addresses the challenge of interpreting dynamic, emerging memes by retrieving real-time web evidence. It outperforms existing zero-shot baselines in both meme understanding and detection tasks.
This ArXiv paper benchmarks LLMs on Solidity smart contracts using zero-shot, CoT, and ToT prompting for error detection and classification on 400 contracts. CoT and ToT raise recall to 95-99% in detection but drop precision. Claude 3 Opus tops error classification with 90.8 Weighted F1 via ToT.
2-year project optimizes live line-haul logistics with MARL 'Fleet Manager' (PPO) for high-level clustering/dispatch and LP 'Dock Worker' for bin packing/TSP. Achieves zero-shot generalization via normalized density maps. Full architecture deep dive shared.
Multi-dimensional zero-shot benchmark evaluates four TSFMs (Chronos, Moirai, TinyTimeMixer) vs. baselines on ERCOT data. Tests context sensitivity, calibration, robustness to shifts like COVID/Winter Storm. Top models hit MASE 0.31; Chronos-2 best calibrated.