Catalog Systems Memory benchmarks & evaluation PROCGEN benchmark suite

PROCGEN benchmark suite

https://github.com/openai/procgen

OpenAI's 16-game procedurally-generated benchmark for evaluating RL generalisation. Unlike Atari (fixed levels memorisable), every PROCGEN episode uses a different procedural seed → test-set generalisation is the headline metric. Released at ICML 2020. Standard benchmark for studying generalisation, exploration, transfer.

Active · Verified 2026-10-28 · fresh as of 2026-10-28 · 31 cited facts, 1 with a third-party source · Suggest a correction

At a glance

Type
Procedurally-generated RL game suite — generalisation eval repo fix
Tier
T3
Section
Memory benchmarks & evaluation
Created
2019-12 (Cobbe et al. arXiv 1912.01588) repo fix
Latest release
not applicable — not OSS fix
License
MIT repo fix
GitHub
980★ Python/C repo fix
Pricing
not applicable — open benchmark; free to use repo fix
Funding
not applicable — not commercial fix
Freshness
2026-10-28 fix
Layer
agent-infra repo fix
Licensing
open-source repo fix
Developer activity
OpenRank 0.32 (12-mo avg 1.01) · 1 participants — 2024-08 third-party fix

Memory taxonomy

storage
n/a
retrieval
n/a
persistence
n/a
update
read-only
unit
n/a
governance
n/a
conflict
n/a

When to use

Optimised for: not applicable - eval dataset, not a system

Anti-fit: not applicable - eval dataset, not a system

Pros & cons

Pros

Procedural seed-train / seed-test split rigorously isolates generalisation from memorisation; lightweight (16 fast games).

Cons

2D platformer style limits transfer to richer 3D / embodied; PROCGEN-shaped overfitting is a known critique.

Claims & capabilities

ICML 2020; 16 procedurally-generated games; standard for RL generalisation research; used by IMPALA, PPG, DAAC, IDAAC follow-ups

Technical surface

API surface
not applicable — eval dataset, not a system
Backend storage
not applicable — eval dataset, not a system
Deployment
not applicable — run locally
Embedding model
not applicable — eval dataset, not a system
Multi-tenancy
not applicable — eval dataset, not a system
MCP
not applicable — benchmark / evaluation harness
A2A
not applicable — benchmark / evaluation harness
OpenTelemetry
not applicable — benchmark / evaluation harness

Similar systems

Other memory benchmarks & evaluation in the catalog, ranked by inbound references.

  • LoCoMo T3

    Conversations spanning up to 35 sessions / 300 turns / ~9K tokens, built from LLM-agent personas grounded on temporal event graphs and verified by human annotators. Tests QA (single-hop, multi-hop, temporal, commonsense, adversarial), event-graph summarisation, multimodal dialogue.

  • GAIA T3

    Benchmark for general AI assistants made of real-world questions that require reasoning, multimodality, web browsing and tool use, with a public leaderboard on Hugging Face. Traction: ~917 likes on the Hugging Face dataset

  • LongBench T3

    Tsinghua THUDM. 21 datasets across 6 categories (single-doc QA, multi-doc QA, summarisation, few-shot ICL, synthetic, code completion) in English + Chinese. ~6.7K words English / 13.4K chars Chinese average. v2 at ACL 2025 with harder, longer instances.

  • LongMemEval T3

    5 core memory abilities — info extraction, multi-session reasoning, temporal reasoning, knowledge updates, abstention — via 500 curated questions in synthetic chat histories. Two scales: LongMemEval-S (115K tokens), LongMemEval-M (up to 1.5M).

  • RULER T3

    NVIDIA. Extends NIAH with multi-needle retrieval, multi-hop entity tracing, aggregation, QA — configurable for arbitrary length / difficulty. Evaluated 17 long-context models on 13 tasks.

  • InfiniteBench (∞Bench) T3

    Tsinghua / OpenBMB. First benchmark with average input >100K tokens. 12 tasks across English + Chinese: long novel QA, code debugging, math in long context, summarisation, retrieval. Realistic documents (novels, code) rather than synthetic filler.

Sources (30 cited facts, 1 third-party)

Each fact links to the page it came from; the badge says who published it (how to read sources).