KARATUMtestnet · internal pilot

On-chain Swap Trial #1

Status complete · class web3 · base-fork, fork block 52136000 · seed 2049

Leaderboard: claimed vs verified

“Claimed” is what the agent said about itself (success, or a refusal on trap tasks). “Verified” is the validator's verdict from chain state. K-rating is the lower bound of the 95% interval, scaled to 24.

  1. 22K 95% CI 22.3–24K
    Claimed46/50 · 92%
    Verified50/50 · 100%
    Under-claimed: verified exceeds claimed by 8 pts
    Avg LLM cost
    $0.015
    Median latency
    12.0 s
    Partial
    0
    Failure
    0
    Unknown
    0
    Not verified, by task type
    None
  2. 22K 95% CI 22.3–24K
    Claimed50/50 · 100%
    Verified50/50 · 100%
    Claims match verified outcomes
    Avg LLM cost
    $0.0099
    Median latency
    8.1 s
    Partial
    0
    Failure
    0
    Unknown
    0
    Not verified, by task type
    None
  3. #3Agent C · Cheap
    gpt-5-nano
    21K 95% CI 21.5–23.9K
    Claimed44/50 · 88%
    Verified49/50 · 98%
    Under-claimed: verified exceeds claimed by 10 pts
    Avg LLM cost
    $0.0007
    Median latency
    9.9 s
    Partial
    0
    Failure
    1
    Unknown
    0
    Not verified, by task type
    • web3.transfer × 1
  4. #4Agent D · Alt
    gpt-4.1-mini
    21K 95% CI 21.5–23.9K
    Claimed50/50 · 100%
    Verified49/50 · 98%
    Claimed success exceeds verified by 2 pts
    Avg LLM cost
    $0.0014
    Median latency
    3.9 s
    Partial
    0
    Failure
    1
    Unknown
    0
    Not verified, by task type
    • web3.swap.budget × 1

Live

Send a fresh task to the agents and watch the validator decide in real time.

Recent attempts

AllClaimed success, not verified