Kw bench benchmark tests real world investigative knowledge tasks
kw-bench: Breaking Frontier Harnesses on Real-World Tasks
Artificial IntelligenceSoftware Engineering
Summary
Investigative knowledge work, like audits or tax checks, requires finding correct answers hidden in messy, incomplete data. Existing AI tests often focus on creating new content, not digging through real data to uncover facts. The authors created kw-bench, a test based on real business scenarios involving complex VAT data across multiple systems. They showed that AI models combined with effective tools can solve these tough problems much faster and more accurately than older methods.
What this means in practice
- •For audit and tax teams: Automate complex financial data reconciliation tasks involving multiple legacy systems using AI with specialized harnesses for faster and reliable outcomes.
- •For regulatory compliance teams: Accelerate investigative reporting and dispute resolution by deploying benchmarks that validate AI tools handling fragmented and inconsistent evidence.
Authors
George Juraj Salapa
Abstract
Each new generation of frontier model pushes the edge of what is possible. Inside a harness -- a loop, a shell, a filesystem, the ability to write and run its own code -- a model can take on increasingly difficult work. A growing body of work treats the coding-agent harness as infrastructure for knowledge work rather than a coding tool, and a new wave of benchmarks tests it there. Most of them are generative: build a financial model, a memo, a spreadsheet from a brief, judged by a rubric or by another model. Few test the expensive part of knowledge work, which is investigative. In audit, tax, due diligence, regulatory reporting, claims or any dispute over numbers, the answer already exists -- hidden in incomplete and inconsistent evidence spread across systems -- and has to be recovered and defended to the cent. We introduce kw-bench, a benchmark built from real engagements specialists described to us. Its first task family is a five-year VAT reconstruction across three generations of SAP -- legacy ECC, a partial S/4HANA migration and a consultant's redesign -- delivered as raw system extracts: page headers, local number formats, partial link tables, no documentation of how anything connects. Grading is deterministic, with no LLM judge. We ran four harnesses -- Claude Code, Codex, OpenCode and micro-cc -- on the same model, DeepSeek v4.1 Flash, for two hours. The two independent, model-agnostic harnesses delivered (OpenCode 0.542, micro-cc 0.528); the two lab harnesses, away from their home model, delivered nothing (0.00 each). What defeated them was inertia: keeping many weak signals alive across a long session. On a frontier model the task collapses. micro-cc and Claude Code on Claude Opus 5.5 scored 0.9986 and 1.000, in 22 and 25 minutes. A weeks-long engagement now costs minutes and a few dollars; below the frontier, the harness decides how far a model gets.