spb/wpai0001_uqo

1% source: experiments/summaries/tier1_taskvector_confirmatory_625c54bd186f/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\midrule6qwen3b $\to$ qwen1.5b & antonym & yes & replace\_donor\_fewshot procrustes L16 & 0.19 & 0.94 & 0.91 & 0.59 & 0.44 & +0.16 [+0.00, +0.34] & 0.69 \\7qwen3b $\to$ qwen1.5b & capital & yes & ridge L22 a1.0 & 0.00 & 1.00 & 0.82 & 0.09 & 0.00 & +0.09 [+0.00, +0.23] & 0.00 \\8qwen3b $\to$ qwen1.5b & morph & no & procrustes L2 a1.0 & 0.00 & 0.30 & 0.00 & 0.00 & 0.00 & +0.00 [+0.00, +0.00] & 0.00 \\9qwen3b $\to$ qwen1.5b & addthree & no & procrustes L2 a1.0 & 0.00 & 0.86 & 0.00 & 0.00 & 0.00 & +0.00 [+0.00, +0.00] & 0.00 \\10\multicolumn{11}{l}{\emph{qwen3b $\to$ qwen1.5b: H6 not supported (0 task(s))}} \\ \addlinespace11qwen1.5b $\to$ qwen3b & antonym & yes & procrustes\_rescaled L21 a1.0 & 0.09 & 0.94 & 0.91 & 0.69 & 0.41 & +0.28 [+0.09, +0.47] & 0.81 \\12qwen1.5b $\to$ qwen3b & capital & yes & procrustes L28 a1.0 & 0.14 & 1.00 & 0.82 & 0.55 & 0.18 & +0.36 [+0.14, +0.59] $\checkmark$ & 0.27 \\13qwen1.5b $\to$ 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 \\14qwen1.5b $\to$ 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 \\15\multicolumn{11}{l}{\emph{qwen1.5b $\to$ qwen3b: H6 partial (1 task(s))}} \\ \addlinespace16llama3b $\to$ qwen3b & antonym & yes & cca L25 a1.0 & 0.09 & 0.94 & 0.91 & 0.81 & 0.31 & +0.50 [+0.31, +0.69] $\checkmark$ & 0.81 \\17llama3b $\to$ qwen3b & capital & yes & cca L18 a1.0 & 0.14 & 1.00 & 0.82 & 0.64 & 0.45 & +0.18 [+0.05, +0.36] & 0.41 \\18llama3b $\to$ 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 \\19llama3b $\to$ 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 \\20\multicolumn{11}{l}{\emph{llama3b $\to$ qwen3b: H6 partial (1 task(s))}} \\ \addlinespace21qwen3b $\to$ llama3b & antonym & yes & procrustes L14 a1.0 & 0.06 & 0.91 & 0.84 & 0.59 & 0.34 & +0.25 [+0.09, +0.41] & 0.38 \\22qwen3b $\to$ llama3b & capital & yes & procrustes L14 a1.0 & 0.09 & 1.00 & 1.00 & 0.09 & 0.14 & -0.05 [-0.18, +0.09] & 0.09 \\23qwen3b $\to$ 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 \\24qwen3b $\to$ llama3b & addthree & yes & procrustes L2 a1.0 & 0.00 & 1.00 & 0.69 & 0.00 & 0.00 & +0.00 [+0.00, +0.00] & 0.00 \\25\multicolumn{11}{l}{\emph{qwen3b $\to$ llama3b: H6 not supported (0 task(s))}} \\ \addlinespace26\bottomrule27\end{tabular}28