list-negative-batches-at-50

IN premise

`list_negative` splits candidate nodes into batches of approximately 50 for LLM classification, verified by the test suite asserting exactly 3 LLM calls for 120 candidates.

Summary

The system processes candidate nodes in groups of about 50 at a time when asking the LLM to identify negative relationships. This batching strategy is confirmed by tests showing that 120 candidates result in exactly 3 LLM calls, which matters because changing the batch size would break test expectations and could affect classification accuracy or API costs.

Dependents

These beliefs depend on this one:

list-negative-is-bounded-and-batch-scalable — The list-negative classification pipeline is both defensively bounded (two-stage keyword + LLM filtering with hallucination rejection and graceful malformed-output handling) and scalably partitioned (fixed batch size of ~50 candidates per LLM call), ensuring predictable resource usage and bounded LLM costs regardless of belief network size.

Details

Source

entries/2026/05/05/tests-test_api.md