mohammadtavakoli78 / BEAM
[ICLR 2026] Beyond a Million Tokens: Benchmarking and Enhancing Long-Term Memory in LLMs
Why this job: says: long term memory; topics: long-term-memory
What it does
BEAM is a comprehensive dataset for evaluating long-term memory in language models. It contains multi-scale conversations (128K, 500K, 1M, and 10M tokens) spanning diverse domains, including general, coding, and mathematical topics, and is designed to assess ten distinct memory abilities. To evaluate LLMs on these abilities, we generate a set of probing questions for each conversation.
Get started
python src/beam/download_dataset.py - Maintenance
- Active
- Last commit 2 months ago
- Stars
- 173
- +6 this week
- Licence
- MIT
- Allows business use
- Downloads for
- None listed
- No release files
- Latest version
- None yet
- No releases
- Built with
- Python
- 100% of the code
Is it looked after?
- Contributors
- 0
- Last 90 days
- Issues
- 19% closed
- 13 still open
- Pull requests
- 0 merged
- 2 waiting
- Releases
- None
- Never published
- No contributing guide
- No security policy
- No code of conduct
More: languages, topics and project record
Languages
- Python 100%
- Shell <1%
Topics
- beam
- beam-benchmark
- benchmark-long-context
- benchmark-long-term-memory
- benchmarking
- light
- light-framework
- llm
- long-context-evaluation
- long-context-evaluation-benchmark
- long-context-llm
- long-context-llms
Project record
- JobAgent memory
- StatusListed
- Commits17
- Forks18
- Watching2
- Stars in 28 days+15
- CreatedOct 28, 2025
- First seen by usNov 7, 2026
- CheckedNov 7, 2026