spb/wpai0001_uqo

1% source: experiments/summaries/tier1_taskvector_confirmatory_532eb68bd5e2/tier1_confirmatory.json (scripts/tier1_analysis.py)2\begin{tabular}{llllcccccll}3\toprule4Pair & task & eligible & selected condition (dev) & zero & few & own & selected & best null control & selected $-$ control [95\% CI] & shuffled-query diag. \\5\midrule6qwen7b $\to$ qwen3b & antonym & yes & replace\_donor\_fewshot ridge L25 & 0.09 & 0.94 & 0.91 & 0.84 & 0.31 & +0.53 [+0.34, +0.72] $\checkmark$ & 0.81 \\7qwen7b $\to$ qwen3b & capital & yes & ridge L14 a1.0 & 0.09 & 1.00 & 0.77 & 0.32 & 0.23 & +0.09 [+0.00, +0.23] & 0.23 \\8qwen7b $\to$ qwen3b & morph & no & procrustes L2 a1.0 & 0.00 & 0.40 & 0.00 & 0.00 & 0.00 & +0.00 [+0.00, +0.00] & 0.00 \\9qwen7b $\to$ qwen3b & addthree & no & procrustes L2 a1.0 & 0.00 & 1.00 & 0.00 & 0.00 & 0.00 & +0.00 [+0.00, +0.00] & 0.00 \\10\multicolumn{11}{l}{\emph{qwen7b $\to$ qwen3b: H6 partial (1 task(s))}} \\ \addlinespace11qwen3b $\to$ qwen7b & antonym & yes & replace\_donor\_fewshot procrustes L19 & 0.25 & 0.97 & 0.72 & 0.75 & 0.56 & +0.19 [+0.03, +0.34] & 0.78 \\12qwen3b $\to$ qwen7b & capital & yes & ridge L11 a1.0 & 0.23 & 1.00 & 0.91 & 0.86 & 0.73 & +0.14 [+0.00, +0.27] & 0.59 \\13qwen3b $\to$ qwen7b & morph & no & procrustes L2 a1.0 & 0.00 & 0.42 & 0.00 & 0.00 & 0.00 & +0.00 [+0.00, +0.00] & 0.00 \\14qwen3b $\to$ qwen7b & addthree & no & procrustes L2 a1.0 & 0.00 & 0.97 & 0.00 & 0.00 & 0.00 & +0.00 [+0.00, +0.00] & 0.00 \\15\multicolumn{11}{l}{\emph{qwen3b $\to$ qwen7b: H6 not supported (0 task(s))}} \\ \addlinespace16qwen7b $\to$ llama3b & antonym & yes & procrustes L14 a1.0 & 0.06 & 0.91 & 0.84 & 0.34 & 0.34 & +0.00 [+0.00, +0.00] & 0.16 \\17qwen7b $\to$ llama3b & capital & yes & procrustes L11 a1.0 & 0.05 & 1.00 & 1.00 & 0.00 & 0.14 & -0.14 [-0.32, +0.00] & 0.05 \\18qwen7b $\to$ llama3b & morph & no & procrustes L2 a1.0 & 0.00 & 0.28 & 0.00 & 0.00 & 0.00 & +0.00 [+0.00, +0.00] & 0.00 \\19qwen7b $\to$ llama3b & addthree & yes & procrustes L2 a1.0 & 0.00 & 1.00 & 0.71 & 0.00 & 0.00 & +0.00 [+0.00, +0.00] & 0.00 \\20\multicolumn{11}{l}{\emph{qwen7b $\to$ llama3b: H6 not supported (0 task(s))}} \\ \addlinespace21\bottomrule22\end{tabular}23