verifiers

Compile real-world Claude Code and Codex trajectories into verified, tradable post-training assets.

Procedural data generators for verifiable reasoning, synthetic pretraining, post-training, evaluation, and RL.

Score the trustworthiness of outputs from any LLM in real-time

🐀 Fuzz your verifier before an RL agent does. Static + dynamic LLM security auditor to detect reward-hacking in RL post-training environments (OpenEnv, verifiers-spec, Gymnasium).

Open Arena: SLM verifiers across observability platforms, dataset and model catalogs, and value scenarios for LLM, agentic and harness evals.

An RL Enviorment for AES Inversion

Adversarial QA for LLM-RL environments: find out what reward an empty answer earns. Model-free, zero API cost.

A verifiers RLM environment for testing whether adaptive recursive search outperforms brittle manual RAG choreography on long synthetic corpora.

Verifiable RL environments for corporate law & governance — deterministic reward, no LLM judge, fully synthetic worlds.

An open reinforcement-learning (RL) environment that trains LLM agents to use the current fact, not the stale one — verifiable reward for temporal fact-currency, built on verifiers / prime-rl (GRPO, LoRA).

Reproducible verifier audits, datasheets, agreement metrics, and release gates

RLVR coding environment for training and evaluating LLM agents on automated code-fixing tasks.

Research proposal for verifier-gated on-policy distillation with explicit evaluation and claim boundaries

Verifiers hello world repo

Minimal verifier environments for testing agents.

RL environments for generating code whose correctness is checked by formal verification; first release: C + ACSL + Frama-C.

Deterministic SRT captioning environment for LLM evaluation and reinforcement learning.

A verifiers RL environment that trains models to propose novel, evidence-grounded, falsifiable hypotheses. Rewards novelty with accountability.

Three small demos on verifiable evaluation: measured retrieval accuracy, a validation loop that bounces bad work, and an end-to-end pipeline. No API key, no GPU.

Linear Algebra Done Right (Axler 4e) as an exactly-graded RL environment for the Prime Intellect Environments Hub. 186 tasks, 25 topics, binary reward from exact arithmetic over the rationals. No LLM judge, no network.