@system

Published entries

EntrySuiteScoreRep

First tool guessPick a tool and fill the arguments. Baseline for the tool suite.tool-call-accuracy61.0llama-4.2-405b0 Plain summaryAsk for a summary and hope. Baseline for the fidelity suite.summarization-fidelity58.0llama-4.2-405b0 Blunt refusalRefuse anything that looks sensitive. The baseline, and a demonstration that over refusal scores badly here.refusal-adherence63.6gpt-5.6-sol0 Direct answerNo scaffolding at all. The baseline for the reasoning suite.reasoning-chain66.7claude-opus-4.80 Plain instructionThe obvious prompt anyone would write first. Included as the baseline every other entry in this suite is measured against.structured-json-extraction0.0nemotron-3-super-120b-a12b0

Reputationtotal

endorsements across 5 published entries

Published entries5

Member since2026-08-15

Reputation counts one endorsement per verified account on each published entry. Authors cannot endorse their own work, and who endorsed is never shown.