SE-Bench

A diagnostic coding test that renames NumPy into a fake library so a score reflects whether an agent internalized new APIs, not old knowledge or hard reasoning.

Also known as: SE-Bench: Benchmarking Self-Evolution with Knowledge Internalization

unassessed

This page is a discovery lead. Nobody has yet assessed it against the catalogue contract, so it carries no disposition. Absence of evidence here is not evidence of staleness.
Categorycoding
Subcategoryknowledge internalization / self-evolution diagnostic
Page statusactive
Metricaccuracy (test-case + AST constraints)
Directionhigher_is_better
Unit%
Dataset size1417
Dataset licenceMIT
PublisherTHUNLP, Tsinghua University

What it measures

SE-Bench checks whether an agent can absorb a new library and later use it with no docs. The authors wrap 268 common NumPy functions as a package named zwc, with nonsense identifiers such as zwc.kocito. Inputs and outputs are ZWCArray objects so the model cannot call NumPy methods on arrays. Training items include the relevant docstring. Test items are ordinary coding problems without that docstring. Single-function tests check recall. Multi-function tests require composing at least three APIs. A base model scores 0% without docs, so failures are meant to be memory failures.

Task format

English problem statement in; the model writes Python that must call zwc APIs, pass hidden tests, and must not import numpy. Training may include API docs; the official test run does not.

Models reporting this benchmark

No model card in ModelSpec reports this benchmark yet.

Data

This page as JSON · Edit on GitHub