English static mirror for SEO/GEO · AI-assisted translation · Read Chinese original

OOLONG Benchmark: Deep Dive into Long-Context Reasoning Limits and 2025–2026 Advances

Forum topic · ✨步子哥 · 2026-01-20

Summary

OOLONG is a 2025 long-context evaluation benchmark released as arXiv:2511.02817 by MIT CSAIL, designed to test true information aggregation and multi-hop reasoning instead of simple retrieval. It contains OOLONG-Synth (synthetic ICL tasks) and OOLONG-Real (Dungeons & Dragons transcripts), with linear tasks and quadratic OOLONG-Pairs requiring all entity pairings. Frontier models in late 2025 scored under 50% accuracy on 128K standard OOLONG and near-zero F1 (0.04%–0.1%) on OOLONG-Pairs, even at 263K tokens, exposing context rot and phase-transition collapse. The Recursive Language Models (RLM) paper (arXiv:2512.24601) used OOLONG as the core benchmark, boosting Pairs F1 from under 0.1% to 58% by loading context into a Python REPL and recursing on code-generated sub-queries. Open datasets, evaluation scripts, and community RLM implementations are available on GitHub and Hugging Face.

Overview

OOLONG (Oolong: Evaluating Long Context Reasoning and Aggregation Capabilities) is one of the most demanding long-context benchmarks released in late 2025. Unlike Needle-in-a-Haystack style retrieval tests, OOLONG forces models to perform true information aggregation and multi-hop reasoning across very long documents, exposing structural weaknesses in current transformer architectures.

Core Design and Difficulty Sources

  • Release: November 4, 2025, arXiv:2511.02817, lead author Andrew Bertsch and colleagues at MIT CSAIL.
  • Two subsets:
  • OOLONG-Synth: Synthetic tasks adapted from classical in-context learning datasets. Strong controlled variables make it suitable for analyzing verifiable aggregation behavior.
  • OOLONG-Real: Built from live Dungeons & Dragons transcript data. Questions require narrative comprehension and multi-character interaction aggregation that cannot be decomposed trivially.
  • Task complexity levels:
  • Linear OOLONG: Sequential line-level analysis, counting, or aggregation (e.g., sum a property across all characters).
  • Quadratic OOLONG-Pairs: The most punishing setting. Requires pairwise processing of every entity in the document (e.g., counting interactions between every character pair). Complexity scales as O(n²), and at 128K+ tokens traditional models collapse.
  • Why it is hard:
  • Most long-context benchmarks are retrieval-based.
  • OOLONG is aggregation-based and demands atomic understanding of every segment before global integration.
  • Attention degrades on very long sequences, producing "context rot" and abrupt phase transitions where capability collapses suddenly.
  • Outputs are numerically and logically verifiable, removing subjective grading ambiguity.
  • Frontier Model Performance

  • On 128K standard OOLONG, several late-2025 frontier models scored below 50% accuracy.
  • On OOLONG-Pairs, baseline models—even those with million-token windows—achieved F1 scores between roughly 0.04% and 0.1%, essentially failing.
  • Performance continued to degrade up to 263K tokens.
  • The result demonstrates that merely enlarging the context window does not guarantee usable long-context reasoning.
  • RLM on OOLONG: A Large Jump

    The Recursive Language Models (RLM) paper (arXiv:2512.24601, December 31, 2025, Alex Zhang et al., MIT CSAIL) used OOLONG as its central benchmark.

  • On OOLONG-Pairs: baseline under 0.1% → RLM reaches approximately 58% F1, a gain of more than 50 percentage points.
  • At 10M+ token scales, RLM maintains high accuracy while summary- or RAG-based approaches either explode in cost or fail outright.
  • Smaller setups such as RLM + GPT-5-mini outperform native GPT-5 on OOLONG by about 114%.
  • Mechanism: long context is loaded into a Python REPL; the model emits code to slice, filter, and recursively sub-query the context, fetching only the necessary fragments.
  • Open Resources and Community Activity (as of January 2026)

  • GitHub: https://github.com/abertsch72/oolong (datasets, evaluation scripts, model outputs).
  • Hugging Face: datasets such as oolongbench/oolong-real are publicly downloadable.
  • Open-source RLM implementations include https://github.com/alexzhang13/rlm and community forks such as https://github.com/ysz/recursive-llm.
  • Community discussion treats OOLONG as a "stress test" for long context; reproductions commonly report near-zero baseline scores on Pairs. RLM is described by some practitioners as a 2026 paradigm shift and is being integrated into agent frameworks for codebase-scale analysis.
  • Implications

    OOLONG challenges the assumption that longer windows equal stronger reasoning. Combined with the neuro-symbolic approach of RLM—precise code-level control plus local sub-model reasoning—it points to a path beyond raw context scaling: teaching models to manage context the way a programmer would.

    Key References

  • OOLONG paper: arXiv:2511.02817
  • Recursive Language Models: arXiv:2512.24601
  • Code: https://github.com/abertsch72/oolong
  • RLM reference implementation: https://github.com/alexzhang13/rlm
  • Dataset: oolongbench/oolong-real on Hugging Face

Tags

#long-context#benchmark#oolong#rlm#reasoning#aggregation#mit-csail#evaluation

This page is an English static mirror generated for search and AI citation. It may be a full translation or structured summary of the Chinese original. Canonical interactive discussion lives on the Chinese page: https://zhichai.net/topic/176415305