Unverified50% confidenceFactExact time
LongVALE focuses on joint vision-audio-language understanding, while UnAV-100 focuses on audio event localization; both are commonly used benchmarks for audio-visual temporal grounding.
1
Sources
50%
Confidence
Long-term
Relevance
10/20/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedThe research team proposed HLR-AVSceneQA, a comprehensive spatial audio-visual evaluation benchmark designed according to the 'Hear-Localize-Reason' hierarchical structure.71% similarUnverifiedMost current audio-visual large language models (AV-LLMs) treat audio as monaural semantic information, and can only recognize sound content but cannot determine the direction of the sound source or its association with objects in the scene.69% similarPartially VerifiedAudio is a time-series signal, and music generation models need to maintain coherence over long time spans; integrating them into a workflow engine primarily designed for images/video requires specialized adaptation in sampling scheduling, memory management, and output format67% similarUnverified音视频联合去噪将音频潜空间与视频潜空间在同一扩散过程中并行处理，音频信号作为条件约束参与每一步去噪迭代66% similarUnverifiedAudio is more sensitive to numerical precision and distortion is easily detected by the ear, so fp32 is chosen for the audio VAE; fp16 is chosen for the video VAE to strike a balance between image quality and VRAM usage.66% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1073806API
curl https://kongchang.com/api/v1/knowledge/claims/1073806MCP
get_claim(id=1073806)