中文版 / Chinese version

LongMemEval Benchmark

LongMemEval is a benchmark for long-term memory over multi-session chat histories. Each item provides a chronologically ordered set of chat sessions between a user and an assistant, followed by a probing question whose answer is only recoverable by reasoning over the user-owned memory. ReMe ingests the sessions into an isolated per-item workspace, answers the question via an agentic (ReAct) mode, and scores the answer with an LLM-as-judge.

Question types include single-session (user / assistant / preference), multi-session reasoning, knowledge update, and temporal reasoning.

Install ReMe and the LongMemEval plugin in editable mode from the repository root:

python -m pip install -e ".[as]"
reme plugins install ./plugins/lme --editable
reme plugins install ./plugins/lme-judge --editable
reme plugins validate lme

The runner explicitly enables the installed lme plugin and combines its defaults with ReMe's built-in benchmark preset. Editable installation keeps changes under plugins/lme visible without reinstalling the plugin. Custom application config paths still work through reme.config and can use extends: benchmark. This directory continues to own the runner, evaluation settings, dataset and outputs. Model credentials use the environment variables declared by the shared benchmark configuration.

1. Get the Dataset

ReMe uses only the cleaned-S split, hosted on HuggingFace: agentscope-ai/ReMe_longmemeval_clean_s_v2. The download script fetches it via the hf-mirror.com mirror; to use a different mirror, modify BASE_URL in download.py.

cd benchmark/longmemeval
python download.py            # saves dataset/longmemeval_s_reme_cleaned.json; skips if already present

Ground truth is embedded in the data file.

2. Run

From the repository root:

python benchmark/longmemeval/run.py
python benchmark/longmemeval/run.py --config benchmark/longmemeval/config.yaml
python benchmark/longmemeval/run.py -q                        # quiet: only eval-level logs
python benchmark/longmemeval/run.py --log-level WARNING       # reduce eval runner logs
python benchmark/longmemeval/run.py --reme-log-level WARNING  # reduce reme internal logs
python benchmark/longmemeval/run.py --eval_only               # reuse existing workspaces, query + judge only

3. Pipeline

1. Load the dataset (ground truth is embedded in the data file).
2. For each item, create an isolated workspace and ingest sessions in chronological order.
3. If a custom application configuration enables auto_dream, trigger it when sessions cross the configured hour (default 23:00). The packaged preset leaves it disabled.
4. Answer each question via agentic (ReAct) mode.
5. Judge the answer (binary yes/no) with the answer_judge job and print per-type accuracy.

4. Key config — benchmark/longmemeval/config.yaml

Key	Meaning
dataset.path	Dataset file to evaluate (e.g. longmemeval_s_reme_cleaned.json); ground truth is included.
dataset.start_index / num_items	Slice of items to evaluate.
dataset.question_types	Filter by question type; empty = all.
dataset.workspace_root	Per-item workspace root (benchmark/longmemeval/workspaces/longmemeval-s).
evaluation.num_workers	0 = auto (cpu-2), 1 = sequential, >1 = parallel.
evaluation.filter_future_sessions	Only ingest sessions with timestamp ≤ question_date.
reme.config	ReMe config used (benchmark).
reme.dream_trigger_hour / dream_scan_days / dream_max_units	Dream triggering behavior.
output.dir	Results directory (benchmark/longmemeval/results).

5. Outputs

Results are JSON files written to output.dir as results_<timestamp>.json, with a per-type accuracy summary also printed to the console. Logging conventions are shared across benchmarks — see the top-level README.

6. Reference Results

cleaned-s

Basic settings

1. Modified auto-memory prompt, auto-dream disabled.
2. All sessions in reme-memory are strictly earlier than the question time.

Results

agentscope==2.0.4.post1, conda reme env, 32 workers, eval-only (reusing prebuilt memory) (2026-08-06, 500 items, total 10.0 min)

Type	Agentic	input tok/q	output tok/q	total tok/q	tool calls/q
knowledge-update	0.910	31,581	589	32,169	2.90
multi-session	0.842	52,837	1,474	54,311	4.21
single-session-assistant	1.000	15,596	279	15,875	1.89
single-session-preference	0.633	36,802	818	37,620	3.60
single-session-user	0.986	27,433	359	27,792	2.60
temporal-reasoning	0.902	62,674	985	63,659	4.97
OVERALL	0.894	43,448	876	44,324	3.69