YAML Metadata Error:Invalid content in eval.yaml.
Check out the documentation for more information.
Show details
✖ Invalid input
→ at evaluation_framework
| name: "GCI-Bench" | |
| description: >- | |
| Glint Clarity Index Bench: measures whether small language models | |
| (1M-100M parameters) allocate attention*gradient saliency toward | |
| causally-relevant sentences over same-domain distractors, via a | |
| white-box internals probe rather than answer-correctness grading. | |
| evaluation_framework: "custom" | |
| tasks: | |
| - id: "gci_bench" | |
| config: "default" | |
| split: "test" | |
| metric: "gciScore" | |
| reference_implementation: "https://huggingface.co/datasets/Glint-Research/GCI_Bench/blob/main/evaluation_harness.py" |