Add BABILong benchmark adapter

[aijadugar]

Summary

This PR adds support for the BABILong benchmark in MemoryBench.

Changes

• Added a BABILong benchmark adapter
• Loads local BABILong data from data/benchmarks/babilong
• Converts dataset samples into the unified benchmark format
• Registers the benchmark in the benchmark registry and CLI
• Added documentation for dataset setup and usage

Dataset

https://huggingface.co/datasets/RMT-team/babilong

Testing

Verified with:

bun run src/index.ts help benchmarks
bun run src/index.ts list-questions -b babilong -l 5

Result:

• Successfully loaded 2000 questions
• All question types (qa1–qa20) are recognized
• Questions, sessions, and ground truth are mapped correctly

A provider ingest was also started successfully. The run stopped due to an external OpenAI API rate limit, not because of the BABILong adapter.

[aijadugar]

Hi @ishaanxgupta and @ved015!
Just a friendly ping whenever you have some time.
Since you're both currently working on the LongMemEval-V2 benchmark, would you also be able to review the BABILong benchmark adapter (#49)
It's ready from my side, and I'd really appreciate any feedback or suggested improvements.

Thanks!