LongBench v2: Der neue Goldstandard zur Messung der Langkontext-Fähigkeiten von KI

GitHub May 2026
⭐ 1163
Source: GitHublong-context AILLM evaluationArchive: May 2026
LongBench v2, die neueste Iteration des führenden Langkontext-Benchmarks des THUDM-Teams der Tsinghua-Universität, wurde auf der ACL 2025 angenommen. Diese neue Version führt schwierigere, realistischere Aufgaben ein, die die Lücke zwischen angegebenen Kontextfenstern und tatsächlicher Modellleistung aufdecken und einen neuen Standard setzen.
The article body is currently shown in English by default. You can generate the full version in this language on demand.

The THUDM team at Tsinghua University has released LongBench v2, a major update to their widely adopted long-context understanding and generation benchmark, with both versions now accepted at ACL 2025 and ACL 2024 respectively. The original LongBench, launched in 2023, quickly became the de facto standard for evaluating how well large language models (LLMs) handle long documents, covering multiple tasks, languages, and length levels. LongBench v2 raises the bar significantly, introducing more challenging tasks that require genuine long-range reasoning—not just retrieval of a single fact. The benchmark now includes synthetic and real-world scenarios that test a model's ability to synthesize information across hundreds of thousands of tokens. This is critical because many models claim context windows of 128K, 200K, or even 1M tokens, but real-world performance often degrades sharply beyond 32K or 64K tokens. LongBench v2's design specifically targets this 'context window gap,' making it an essential tool for researchers and companies alike. The project's GitHub repository has garnered over 1,100 stars, reflecting strong community interest. For AINews, this benchmark is not just an academic exercise—it is a litmus test for the next generation of AI products, from document analysis agents to codebase-level coding assistants. The results from LongBench v2 will directly influence which models enterprises trust for mission-critical long-context applications.

Technical Deep Dive

LongBench v2 is not merely a larger version of its predecessor; it represents a fundamental rethinking of how to evaluate long-context capabilities. The original LongBench covered 21 tasks across six categories: single-document QA, multi-document QA, summarization, few-shot learning, synthetic tasks, and code completion. It used a mix of English and Chinese data, with lengths ranging from 5K to 15K tokens. While useful, this setup had a critical flaw: many tasks could be solved with simple retrieval (e.g., finding a needle in a haystack).

LongBench v2 addresses this by introducing multi-hop reasoning over long contexts. For example, a task might require the model to read a 100K-token novel and then answer a question that requires synthesizing information from three different chapters, each separated by tens of thousands of tokens. This is fundamentally harder than single-fact retrieval. The benchmark also includes:

- Synthetic 'Distractor' Tasks: Inserting irrelevant but plausible information to test whether the model can ignore noise.
- Cross-Lingual Long-Context: Tasks that require understanding a document in one language and answering in another, testing both context length and language transfer.
- Multi-Turn Long Conversations: Simulating a long chat history (e.g., 50+ turns) and testing the model's ability to recall and use information from early turns.

The engineering approach behind LongBench v2 is also noteworthy. The THUDM team developed a dynamic length sampling method to ensure that models are tested at multiple length intervals (e.g., 32K, 64K, 128K, 256K) rather than just one fixed length. This allows for a granular performance curve, revealing exactly where a model's performance begins to degrade.

Benchmark Comparison: LongBench v2 vs. Other Long-Context Benchmarks

| Benchmark | Max Length | Task Types | Multi-Hop? | Cross-Lingual? | Synthetic Noise? |
|---|---|---|---|---|---|
| LongBench (v1) | 15K tokens | 21 tasks (QA, sum., code) | Limited | Yes (EN/ZH) | No |
| LongBench v2 | 256K+ tokens | 30+ tasks (multi-hop, distractors) | Yes | Yes (EN/ZH/FR/ES) | Yes |
| RULER (2024) | 128K tokens | Synthetic needle/haystack variants | No | No | Yes |
| L-Eval | 32K tokens | 18 tasks (QA, sum.) | No | No | No |
| HELMET | 100K tokens | 7 tasks (QA, retrieval) | Limited | No | No |

Data Takeaway: LongBench v2 is the only benchmark that combines multi-hop reasoning, cross-lingual tasks, synthetic noise, and dynamic length sampling. This makes it significantly harder and more realistic than alternatives like RULER, which only tests simple retrieval, or L-Eval, which caps at 32K tokens. The inclusion of synthetic noise is particularly important because real-world documents always contain irrelevant information.

For developers, the LongBench evaluation suite is available on GitHub as an open-source Python package. It integrates with popular model-serving frameworks like vLLM and Hugging Face Transformers, making it easy to run evaluations on custom models. The repository also includes a leaderboard, which has become a key reference point for the community.

Key Players & Case Studies

The THUDM team, led by Professor Jie Tang and senior researcher Zhiyong Wu, has a strong track record in open-source AI. They are also the creators of the ChatGLM series of models, which have been widely adopted in China and globally. The development of LongBench v2 is a natural extension of their work on long-context models; they needed a rigorous benchmark to validate their own models' capabilities.

Competing Benchmarks and Their Sponsors:

| Benchmark | Developer/Sponsor | Focus | Strengths | Weaknesses |
|---|---|---|---|---|
| LongBench v2 | THUDM (Tsinghua) | Realistic multi-hop reasoning | Hardest tasks, cross-lingual, noise | Smaller community than HELMET |
| RULER | Google Research | Synthetic retrieval | Simple, reproducible | Too easy; doesn't test reasoning |
| HELMET | Stanford CRFM | Long-context QA | Good for retrieval tasks | Limited task diversity |
| L-Eval | Shanghai AI Lab | Summarization & QA | Clean dataset | Short max length (32K) |
| SCROLLS | Allen AI | Long-document QA | Real-world documents | Outdated; max 10K tokens |

Data Takeaway: The landscape is fragmented, but LongBench v2 is the only benchmark that directly challenges models to do more than retrieve a single fact. Google's RULER, while popular, is increasingly seen as insufficient because models can achieve high scores on it without demonstrating genuine long-context understanding. This has led to a situation where a model might score 95% on RULER but fail on LongBench v2's multi-hop tasks.

Case Study: OpenAI's GPT-4 Turbo vs. LongBench v2

When GPT-4 Turbo was released with a 128K context window, initial tests on LongBench v1 showed strong performance. However, early results on LongBench v2 (shared by researchers on social media) revealed a significant drop in accuracy on multi-hop tasks when the context exceeded 64K tokens. This suggests that while GPT-4 Turbo can *retrieve* information from 128K tokens, its ability to *reason* across that distance is limited. This finding has direct implications for use cases like legal document analysis or codebase-wide refactoring, where multi-hop reasoning is essential.

Case Study: Anthropic's Claude 3.5 Sonnet

Claude 3.5 Sonnet, with its 200K context window, has performed relatively well on LongBench v2, particularly on tasks involving long conversations and summarization. Anthropic has explicitly designed its models to maintain coherence over long contexts, and the LongBench v2 results validate this strategy. However, the model still struggles with cross-lingual tasks, where performance drops by 15-20% compared to English-only tasks.

Industry Impact & Market Dynamics

LongBench v2 is reshaping the competitive landscape for LLMs in several ways:

1. Exposing the 'Context Window Gap': Many companies market their models based on maximum context window size (e.g., 1M tokens for Gemini 1.5 Pro). LongBench v2 provides a standardized way to measure *effective* context length—the length at which performance remains acceptable. This is a much more meaningful metric for enterprise buyers.

2. Driving Investment in Long-Context R&D: The benchmark's difficulty is pushing companies to invest in new architectures. For example, Meta's research on 'Ring Attention' and 'Striped Attention' is directly motivated by the need to scale context windows without quadratic compute costs. Startups like Contextual AI and Fixie are also building products specifically optimized for long-context tasks, and they use LongBench v2 as a key validation tool.

3. Impact on Enterprise Adoption: Enterprises are increasingly looking to deploy LLMs for tasks like contract analysis, codebase understanding, and customer support history analysis. A model that scores well on LongBench v2 is more likely to be trusted for these use cases. Conversely, a model that fails on LongBench v2 will face skepticism, regardless of its marketing claims.

Market Size and Growth Data:

| Year | Long-Context LLM Market (Est.) | Key Drivers | LongBench v2 Adoption (Papers Citing) |
|---|---|---|---|
| 2023 | $2.5B | GPT-4 release, early enterprise pilots | 0 (not released) |
| 2024 | $6.8B | Claude 3, Gemini 1.5, open-source models | 45+ |
| 2025 (Proj.) | $15B+ | Agentic workflows, code assistants, legal AI | 200+ (est.) |

Data Takeaway: The long-context LLM market is growing rapidly, and LongBench v2 is becoming the de facto standard for evaluating these models. The number of papers citing LongBench has grown exponentially, indicating its central role in the research ecosystem.

Risks, Limitations & Open Questions

Despite its strengths, LongBench v2 has limitations:

- Synthetic Task Bias: Some tasks are synthetic (e.g., inserting random numbers into a long document). While useful for controlled testing, they may not reflect real-world performance where context is naturally structured.
- Language Coverage: While LongBench v2 adds French and Spanish, it still lacks coverage for many high-demand languages like Japanese, Arabic, and Hindi. This limits its utility for global enterprises.
- Evaluation Cost: Running a full LongBench v2 evaluation on a 256K-token context can be expensive, requiring significant GPU time. This may disadvantage smaller teams or startups.
- Gaming the Benchmark: As with any benchmark, there is a risk that models will be over-optimized for LongBench v2 tasks, leading to a 'Goodhart's Law' effect where the benchmark ceases to be a reliable measure of general capability.
- Lack of Temporal Reasoning: LongBench v2 does not test a model's ability to handle time-dependent information (e.g., understanding that an event in chapter 1 precedes an event in chapter 5). This is a key capability for tasks like financial analysis or historical research.

AINews Verdict & Predictions

LongBench v2 is the most important long-context benchmark to date, and its acceptance at ACL 2025 solidifies its status as the gold standard. We predict the following:

1. Within 12 months, every major LLM provider will publish LongBench v2 scores as a key performance metric. OpenAI, Google, Anthropic, and Meta will all reference it in their technical reports, just as they currently do with MMLU and HumanEval.

2. The 'context window arms race' will shift from maximum length to effective length. Companies will stop marketing 1M-token context windows and instead focus on 'LongBench v2 Effective Score at 128K' or similar metrics.

3. New architectures will emerge specifically to excel on LongBench v2. We expect to see more research into hierarchical attention, memory-augmented transformers, and retrieval-augmented generation (RAG) systems that can handle ultra-long contexts efficiently.

4. The biggest winner will be Anthropic, whose Claude models already perform well on long-context tasks. OpenAI and Google will need to catch up, which may require significant architectural changes.

5. The biggest loser will be models that rely on simple retrieval tricks. Any model that cannot perform multi-hop reasoning over long contexts will be exposed as inadequate for enterprise use.

What to watch next: The release of LongBench v3 (likely in 2026) will probably include temporal reasoning, more languages, and tasks that require the model to *generate* long-form content (e.g., writing a 50-page report based on 200 pages of source material). The THUDM team has set a high bar, and the industry will be watching closely.

More from GitHub

UntitledEnvoy proxy, the backbone of modern service meshes and API gateways, faces constant performance pressure from feature adUntitledEnvoy Proxy, a high-performance L7 proxy and communication bus originally built at Lyft, has evolved into the foundationUntitledThe SGLang project has quietly become a critical piece of infrastructure for running large language models efficiently. Open source hub2270 indexed articles from GitHub

Related topics

long-context AI24 related articlesLLM evaluation27 related articles

Archive

May 20262931 published articles

Further Reading

AlpacaEval: The Open-Source Benchmark That's Reshaping LLM EvaluationStanford's AlpacaEval offers a cheap, fast, and human-validated way to evaluate instruction-following language models. TFlash Linear Attention: Die Open-Source-Bibliothek, die KI-Modelle mit langem Kontext neu gestaltetEine neue Open-Source-Bibliothek namens Flash Linear Attention erweitert die Grenzen dessen, wie lange KI-Modelle denkenHumanEval: Wie OpenAIs Code-Benchmark die Bewertung von KI-Programmierung neu definiertOpenAIs HumanEval-Benchmark hat grundlegend verändert, wie die KI-Community Code-Generierungsmodelle bewertet. Durch dieEvalPlus: Der rigorose Benchmark, der versteckte Mängel in der KI-Codegeneration aufdecktEin neuer Benchmark namens EvalPlus verändert grundlegend, wie wir die Programmierfähigkeiten großer Sprachmodelle messe

常见问题

GitHub 热点“LongBench v2: The New Gold Standard for Measuring AI's Long-Context Prowess”主要讲了什么?

The THUDM team at Tsinghua University has released LongBench v2, a major update to their widely adopted long-context understanding and generation benchmark, with both versions now…

这个 GitHub 项目在“How to run LongBench v2 evaluation on a custom model”上为什么会引发关注?

LongBench v2 is not merely a larger version of its predecessor; it represents a fundamental rethinking of how to evaluate long-context capabilities. The original LongBench covered 21 tasks across six categories: single-d…

从“LongBench v2 vs RULER benchmark comparison for 2025”看,这个 GitHub 项目的热度表现如何?

当前相关 GitHub 项目总星标约为 1163,近一日增长约为 0,这说明它在开源社区具有较强讨论度和扩散能力。