APEX-Agents · Law

World425_AVK_01

0/1Fail

APEX-Agents task World425_AVK_01 in AI Agents for Tax Due Diligence. Compare dual-harness agent runs across models — rubric criteria, scores, and public traces.

AI Agents for Tax Due DiligenceLaw World 425Dual harnessGrader: rubric

Task IDtask_b78c4510be784e6a8b8f0394aafd785d

WorldLaw World 425

Expected outputmessage_in_console

Published runs7 models · dual config

Task prompt

What the agent was asked to do

Summit Filing Solutions, Inc. bought a building in 2018 for $350k. It spent $250k on the buildout. It sold the building on June 20, 2022 for $1.2M. Review the diligence memo and current cap table. State the amount Laura Kensington owed in federal income taxes attributable to this transaction if her taxable income was $220k in 2022 (assuming no other deductions or credits, or any retroactive changes to the tax treatment). Write me a message, and give the exact dollar amount.

Published trajectories

Agent runs on this task

Curated dual-harness runs (parsed + original sandbox). Best scored run per model.

Model	Harness	Score	Result	Links
GPT-5.5showcase	dual	0/1	Fail	Share page Public trace
fireworks models Kimi K2	dual	0/1	Fail	Share page Public trace
Gemini 3 Flash	dual	0/1	Fail	Share page Public trace
Gemini 3.1 Pro	dual	0/1	Fail	Share page Public trace
GPT-5.4	dual	0/1	Fail	Share page Public trace
GPT-5.4 mini	dual	0/1	Fail	Share page Public trace
GPT-5.4 nano	dual	0/1	Fail	Share page Public trace

Grading rubric

Criteria and grader verdict (showcase run)

States the amount Laura Kensington owed in federal income tax attributable to this transaction is $18,486
Fail
Evidence: <TEXT_RESPONSE> states “Laura Kensington owed **$0**” and “Exact federal income tax owed by Laura attributable to the transaction: $0.” Assessment: Criterion requires stating the amount owed is $18,486; the response gives $0 instead, so it fails.