Unverified50% confidenceBenchmarkExact time
OmniGround reaches 37.5 mIoU on the LongVALE benchmark and 35.7 mIoU on the UnAV-100 benchmark, both achieving SOTA levels for omni-modal temporal grounding.
1
Sources
50%
Confidence
Short-term (~14 days)
Relevance
10/20/2026
First Seen
Valid until: 11/3/2026
Sources
Related Entities
Related Claims
UnverifiedThe 35B model can achieve 45 to 60 tokens per second on a 4090-class GPU.60% similarUnverifiedThe context length of mainstream open-source multimodal models is typically between 4K and 32K tokens: LLaVA-1.5 is 4,096 tokens, Qwen-VL is 8,192 tokens, and InternVL2 extends to 32K.59% similarUnverifiedThe 5M model's configuration includes: 21 quantile nodes from 0.01 to 0.99, context length 32768, and patch size 32.59% similarUnverifiedThe RTX 4090 has a theoretical FP16 compute performance of 330 TFLOPS, and the actual Decode speed for a 7B FP16 model is typically 50–60 tokens/s.58% similarUnverifiedThe Youtu-Parsing-Omni model has 5.33B parameters and about 10.7 GB of weights (BF16).58% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1073804API
curl https://kongchang.com/api/v1/knowledge/claims/1073804MCP
get_claim(id=1073804)