LongBench
A benchmark suite for evaluating the long-context understanding capabilities of large language models, covering tasks such as question answering and summarization.
Timeline (last 90 days)
On LongBench v1/v2 and MemoryAgentBench, NoteBank also reported improvements, covering everything from three 1.3B research models up to Mamba-2-8B and a trained Qwen3.5-9B.
Latent Compaction demonstrates advantages on benchmarks such as RULER, LongBench, and Phonebook Lookup.
The bottleneck in agent memory lies in the write path, yet most memory benchmarks (such as LongBench and Needle-in-a-Haystack) only test long-context retrieval and reasoning, implicitly assuming that memory has already been written correctly.
Benchmark tests commonly used to evaluate long-memory systems include RULER, NIAH (Needle-in-a-Haystack), and LongBench.
In LongBench testing, since no tasks require sequential access, KVFetch's temporal recall channel remains dormant and introduces no additional overhead.
LongBench is a widely used comprehensive benchmark in the field of long-context understanding, covering single-document QA, multi-document QA, summarization, few-shot learning, code completion, and synthetic tasks, with document lengths generally exceeding 8K tokens and some exceeding 32K tokens.
AttSVD's evaluation covers one agent benchmark and the complete LongBench long-context evaluation suite.
Benchmarks commonly used by the open-source community (such as MMLU, HumanEval, MATH, and LongBench) carry risks of data contamination.
All Facts (8)
On LongBench v1/v2 and MemoryAgentBench, NoteBank also reported improvements, covering everything from three 1.3B research models up to Mamba-2-8B and a trained Qwen3.5-9B.
50%UnverifiedLatent Compaction demonstrates advantages on benchmarks such as RULER, LongBench, and Phonebook Lookup.
50%UnverifiedThe bottleneck in agent memory lies in the write path, yet most memory benchmarks (such as LongBench and Needle-in-a-Haystack) only test long-context retrieval and reasoning, implicitly assuming that memory has already been written correctly.
50%UnverifiedBenchmark tests commonly used to evaluate long-memory systems include RULER, NIAH (Needle-in-a-Haystack), and LongBench.
50%UnverifiedIn LongBench testing, since no tasks require sequential access, KVFetch's temporal recall channel remains dormant and introduces no additional overhead.
50%UnverifiedLongBench is a widely used comprehensive benchmark in the field of long-context understanding, covering single-document QA, multi-document QA, summarization, few-shot learning, code completion, and synthetic tasks, with document lengths generally exceeding 8K tokens and some exceeding 32K tokens.
50%UnverifiedAttSVD's evaluation covers one agent benchmark and the complete LongBench long-context evaluation suite.
50%UnverifiedBenchmarks commonly used by the open-source community (such as MMLU, HumanEval, MATH, and LongBench) carry risks of data contamination.
50%