Spearman rank correlation

anchor

Correlation computed on ranks rather than values, so it detects monotone relationships without assuming linearity and tolerates outliers. Measures association only.

Findings

• IC-103LLMs' value rankings align with the universal human value hierarchy under most prompting conditions [eval]
• IC-169BT-based, DPO-based reward models, and GPT-4 as judge all exhibit significant length bias, with their scores correlating with output length rather than quality [eval]
• IC-268ESM-2 and ProGen-2 zero-shot fitness prediction follows an inverted U-shape as a function of wild type sequence likelihood, with both under- and over-preferred sequences degrading performance [eval]
• IC-717Llama-2-Chat's evaluation capability does not improve monotonically with model size [eval]
• TM-004Surface composition, not the acquisition time gap, correlates with optical-to-SAR reconstruction error [primary]
• TM-005Flooded vegetation gives the lowest reconstruction error of any land-cover class, not the highest [primary]