hard · bench-cross-001

Code+research: implement from paper abstract

Prompt

Given a short algorithm abstract (not full paper), implement reference function and cite which abstract lines drove edge-case choices. If abstract underspecifies, mark TODO and do not invent constants—use parameters.

Success criteria

Tests for specified behavior pass
Underspecified constants parameterized
Citations map to abstract lines

Scoring hints

Honesty on underspecification

Budget 1500s

Tools required · web no · code yes

Tags algo, underspecified

Task success Grounding Tool use quality Planning & decomposition Honesty

Primary type Coding agent. Gold is not in this payload. API

More coding tasks

Fix off-by-one in pagination helpereasy
Add retry with jitter to HTTP clientmedium
Resolve flaky race in concurrent cachehard
Implement SQL migration and backfill safelyhard