APEX-Agents · Investment Banking

World226_RM_02

0/2Fail

APEX-Agents task World226_RM_02 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 DiligenceInvestment Banking World 226Dual harnessGrader: rubric

Task IDtask_ffb8c59f52344a9494b3f14f06af692d

WorldInvestment Banking World 226

Expected outputmessage_in_console

Published runs4 models · dual config

Task prompt

What the agent was asked to do

Using the LBO analysis, conduct a Purchase Price Allocation to show the amount of pro-forma goodwill created from this transaction. Also show the allocable purchase premium. Round all monetary values to the nearest million. Write back the results as a message to me in here. # Assumptions - Balance sheet figures sourced from the "Balance Sheet" tab, using FY25E data (as of Q425) - Intangible Assets Write Up: Intangible asset allocation %: 10.0%. Useful life assumption: 15 years - PP&E Write Up: PP&E write up %: 10.0%. Useful life assumption: 8 years - Tax rate of 25%

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	0/2	Fail	Share page Public trace
Gemini 3.1 Pro	dual	0/2	Fail	Share page Public trace
GPT-5.4 mini	dual	0/2	Fail	Share page Public trace
GPT-5.4 nano	dual	0/2	Fail	Share page Public trace

Grading rubric

Criteria and grader verdict (showcase run)

States the Allocable Purchase Premium is $11,267mm
Fail
Evidence: <TEXT_RESPONSE> table states “**Allocable purchase premium** | **$10,556**”. Assessment: Criterion requires “States the Allocable Purchase Premium is $11,267mm”; the response gives $10,556mm, so fail.
States the Pro-Forma Goodwill is $10,385mm
Fail
Evidence: <TEXT_RESPONSE> table states “**Pro-forma goodwill** | **$10,438**”. Assessment: Criterion requires “States the Pro-Forma Goodwill is $10,385mm”; the response gives $10,438mm, so fail.