Unverified50% confidenceBenchmarkExact time
Broken down by task family in JevBench, the accuracy was 0/15 for temporal and numerical questions and 4/19 for long-horizon strategic judgment.
1
Sources
50%
Confidence
Short-term (~14 days)
Relevance
10/18/2026
First Seen
Valid until: 11/1/2026
Sources
Related Entities
Related Claims
UnverifiedResearch shows that 30B-parameter models have a success rate 15–30 percentage points lower than GPT-4/Claude-3.5 on complex reasoning tasks.62% similarUnverifiedOn the 1,302 expert-annotated trajectories of AgentRewardBench, JEV-27B-VL reports an AUROC of 0.91, with precision 78.4 and recall 70.2 at a threshold of 0.5.62% similarUnverifiedIn this test, Base44 scored 100, went from prompt to launch in about 3.5 minutes (the fastest of the four), required 0 fix prompts, and consumed 1 of 25 messages on the free tier.62% similarUnverifiedAcross a range of question sizes spanning 256x, the error judgment rate of a single run is almost unchanged: 0.531 at T=16 and 0.534 at T=4096.61% similarUnverifiedIn the bet-sizing adjustment experiment, Jev's bet-size difference against opponents who fold to large bets was −54 percentage points, Gemini's was −33 percentage points, luna's was +16 percentage points, and Haiku's was +1 percentage point.61% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1064024API
curl https://kongchang.com/api/v1/knowledge/claims/1064024MCP
get_claim(id=1064024)