Backdoor Ifeval

Fresh

Reward hacking with deterministic IF constraints

Type: RL Env
Publisher: Prime
Capabilities: Instruction Following
Tags: Reward Hacking Backdoor
License: unknown
Size: v0.3.0
Published: Apr 2026
Updated: May 2026
Canonical: app.primeintellect.ai/dashboard/environments/prime/backdoor-ifeval

Cite

Notes

Only stored in your browser.

Attribution

README: api.primeintellect.ai/api/v1/environmentshub/prime/backdoor-ifeval/@0.3.0/inspect

Attribution policy →

Evals this tool implements

Same problem set, this tool's harness. Run it to score a model on the test.

IFEval

Google DeepMind

500 prompts with verifiable instruction-following constraints (word counts, casing, JSON format) checked by deterministic rules - no LLM judge needed.

ActiveInstruction Following