Unverified50% confidenceBenchmarkExact time
With SANA-1.6B as the base model and ImageReward as the reward model, SatisDive improved the worst candidate reward by up to 0.70 compared to FK steering.
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
10/5/2026
First Seen
Valid until: 1/3/2027
Sources
Related Entities
Related Claims
UnverifiedOn the Pick-a-Pic dataset using DreamSim as the diversity metric, with FLUX.1-dev as the base model and HPSv3 as the reward model, SatisDive improved the worst candidate reward by up to 0.43 compared to FK steering.73% similarUnverifiedSWE-2's benchmark score differs from that of competitor Fable 5.1 by less than one percentage point, but its cost is 64% lower.60% similarUnverifiedThe Top3 configuration, which uses the fewest metrics, achieved the highest average performance, indicating that fewer metrics actually yield higher performance under this experimental setting.56% similarUnverifiedIn the benchmark, the Dyna-Q algorithm obtained an average of approximately 0.62 badges.55% similarUnverifiedThe same DeepSeek v1.1 model scored between 69.8% and 74.2% on Terminal Bench 2.0 under different harnesses.54% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/975684API
curl https://kongchang.com/api/v1/knowledge/claims/975684MCP
get_claim(id=975684)