Unverified50% confidenceBenchmarkExact time
When trained with the full vocabulary, SL-Ground further reaches an AP of 86.0 at the model granularity.
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
10/20/2026
First Seen
Valid until: 1/18/2027
Sources
Related Entities
Related Claims
UnverifiedIn the AD training stage, the wav2vec 2.0 backbone is frozen, the two task vectors are injected into each Transformer layer with per-layer learnable merging coefficients, and the entire training process optimizes only these merging coefficients and one linear classification head, with trainable parameters under 0.1% of full fine-tuning.58% similarUnverifiedIn the first round of optimization, the changes proposed by the Agent focused on directional convolutions, deep supervision, and training strategies, improving mIoU from 0.817 to 0.82.56% similarUnverifiedWhen handling highly sensitive tasks such as training frontier large models, Opus 5.5 will automatically fall back to a weaker version.55% similarUnverifiedARC-AGI-2, as an upgraded benchmark, raises the difficulty threshold and deliberately circumvents the shortcut of large language models memorizing answers through large-scale pretraining data.55% similarUnverifiedThe classifier trained entirely on synthetic images achieved an AUC-ROC of 0.802 on the real test set, while the baseline trained only on real data achieved 0.760.55% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1074598API
curl https://kongchang.com/api/v1/knowledge/claims/1074598MCP
get_claim(id=1074598)