A sample of TensorBench tasks. Each one names a real piece of functionality the codebase doesn't yet have — the agent has to implement it through the full sparse-tensor compilation pipeline and write tests that exercise the new behavior. The full set of 199 tasks ships with the dataset.
Implement element-wise multiplication (`__mul__`) on `STensor` through the full CIN compilation pipeline, following the same pattern as the existing `__add__` implementation in `stensor.
Add a `transpose()` method and `.
Implement `sum(axis=None)` on `STensor` and as a standalone function in `ops.
Implement `sddmm(S, A, B)` as a first-class operation in `ops.
Add automatic differentiation support for scorch's core operations by implementing custom `torch.
Add compiler-level support for unary operations on sparse tensors.
Want the full dataset?
The complete datasets/tensorbench.json file is in the GitHub repo. Each task object specifies the base commit, container build steps, test command, and the natural-language prompt the agent receives.