| Category | \nSet Name | \n\n \n\n\nSource\n \n | \n\n \n\n\nNumber\n \n | \n
| \n\n\nTool-Using\n | \nTool-Planning | \nAugmented | \n4.4K | \n
| Tool-Calling | \nAugmented | \n4.4K | \n|
| \n\n\nCode\n\nGeneration\n | \nPython-Simple | \nNew | \n2.0K | \n
| Python-Specific | \nNew | \n2.0K | \n|
| Math | \nAugmented | \n2.5K | \n|
| Algorithm | \nGithub | \n2.3K | \n|
| LeetCode | \nLeetCode | \n0.8K | \n|
| Rectification | \nSources Above | \n1.6K | \n|
| \n\n\nTotal\n | \n- | \n- | \n20.0K | \n
| Category | \nTask Name | \n
| \n\n\nMathematics\n | \nArithmetic, Matrix Shape, | \n
| Chinese Remainder | \n|
| \n\n\nCommon Sense\n | \nDate Understanding, Navigate | \n
| \n\n\nLogical Reasoning\n | \nDyck Language, Boolean Expression | \n
| \n\n\nDecomposition\n | \nTracking Shuffled Objects | \n
| Task | \nArith. | \n\n \n\n\nDate U.\n \n | \n\n \n\n\nMatrix S.\n \n | \n\n \n\n\nNavigate\n \n | \n\n \n\n\nChinese R.\n \n | \n\n \n\n\nDyck L.\n \n | \n\n \n\n\nBoolean E.\n \n | \n\n \n\n\nTracking S.\n \n | \n\n \n\n\nAverage\n \n | \n
| \n\n\nNum. of Tools\n | \n5 | \n3 | \n5 | \n2 | \n2 | \n4 | \n2 | \n4 | \n3.38 | \n
| \n\n\nVanilla\n | \n77.78 | \n68.67 | \n40.90 | \n65.16 | \n0.0 | \n19.40 | \n80.70 | \n23.67 | \n47.03 | \n
| \n\n\nCoT\n | \n79.44 | \n68.67 | \n80.46 | \n87.96 | \n0.0 | \n19.42 | \n75.88 | \n40.78 | \n56.58 | \n
| \n\n\nCoS\n | \n100.00 | \n69.28 | \n93.67 | \n85.30 | \n95.14 | \n52.46 | \n97.37 | \n99.11 | \n86.54 | \n
| \n\n\nCoS-Planning\n | \n100.00 | \n66.16 | \n95.18 | \n94.78 | \n100.00 | \n74.58 | \n95.39 | \n99.85 | \n90.74 | \n
| \n\n\nCoS-Calling\n | \n100.00 | \n90.96 | \n97.44 | \n88.44 | \n95.67 | \n98.55 | \n93.42 | \n100.00 | \n95.56 | \n
| Method | \nModel | \nArith. | \n\n \n\n\nDate U.\n \n | \n\n \n\n\nMatrix S.\n \n | \n\n \n\n\nNavigate\n \n | \n\n \n\n\nChinese R.\n \n | \n\n \n\n\nDyck L.\n \n | \n\n \n\n\nBoolean E.\n \n | \n\n \n\n\nTracking S.\n \n | \n
| \n\n\nCoT\n\n(Prompting\n\nw/ demo)\n | \nAlpaca | \n19.89 | \n39.76 | \n5.62 | \n47.11 | \n0.0 | \n0.0 | \n57.46 | \n0.44 | \n
| LLaMA-7B | \n39.44 | \n33.73 | \n12.58 | \n39.70 | \n0.0 | \n2.90 | \n50.44 | \n14.22 | \n|
| ChatGPT | \n79.44 | \n68.67 | \n80.46 | \n87.96 | \n0.0 | \n19.42 | \n75.88 | \n40.78 | \n|
| \n\n\nCoT\n (Tuned) | \nLLaMA-CoT | \n50.44 | \n49.40 | \n70.82 | \n71.64 | \n0.0 | \n35.27 | \n62.72 | \n28.44 | \n
| \n\n\nCoS\n\n(Prompting\n\nw/ demo)\n | \nAlpaca | \n17.78 | \n7.83 | \n3.00 | \n48.60 | \n7.56 | \n1.00 | \n94.74 | \n6.78 | \n
| LLaMA-7B | \n55.89 | \n17.47 | \n10.65 | \n45.90 | \n23.80 | \n35.83 | \n99.12 | \n0.67 | \n|
| ChatGPT | \n100.00 | \n69.28 | \n93.67 | \n85.30 | \n95.14 | \n52.46 | \n97.37 | \n99.11 | \n|
| \n\n\nCoS\n (Tuned) | \nLLaMA-CoS | \n100.00 | \n74.10 | \n91.01 | \n99.56 | \n95.44 | \n98.21 | \n100.00 | \n99.56 | \n
| Method | \nModel | \nArith. | \n\n \n\n\nDate U.\n \n | \n\n \n\n\nMatrix S.\n \n | \n\n \n\n\nNavigate\n \n | \n\n \n\n\nChinese R.\n \n | \n\n \n\n\nDyck L.\n \n | \n\n \n\n\nBoolean E.\n \n | \n\n \n\n\nTracking S.\n \n | \n
| \n\n\nCoS-Whole\n | \nLLaMA-CoS | \n100.00 | \n74.10 | \n91.01 | \n99.56 | \n95.44 | \n98.21 | \n100.00 | \n99.56 | \n
| \n\n\nCoS-Planning\n\n(Prompting\n\nw/ demo)\n | \nAlpaca | \n18.22 | \n27.41 | \n24.15 | \n77.16 | \n100.00 | \n76.3 | \n97.59 | \n99.37 | \n
| LLaMA-7B | \n74.11 | \n27.71 | \n25.02 | \n77.16 | \n100.00 | \n93.80 | \n97.59 | \n100.00 | \n|
| ChatGPT | \n100.00 | \n66.16 | \n95.18 | \n94.78 | \n100.00 | \n74.58 | \n95.39 | \n99.85 | \n|
| \n\n\nCoS-Planning\n | \nLLaMA-CoS | \n100.00 | \n85.84 | \n89.62 | \n97.14 | \n100.00 | \n99.19 | \n97.59 | \n100.00 | \n
| \n\n\nCoS-Calling\n\n(Prompting\n\nw/ demo)\n | \nAlpaca | \n99.44 | \n24.70 | \n30.08 | \n48.60 | \n17.97 | \n1.56 | \n89.91 | \n6.78 | \n
| LLaMA-7B | \n74.70 | \n51.20 | \n55.49 | \n43.77 | \n24.81 | \n25.67 | \n94.30 | \n1.56 | \n|
| ChatGPT | \n100.00 | \n90.96 | \n97.44 | \n88.44 | \n95.67 | \n98.55 | \n93.42 | \n100.00 | \n|
| \n\n\nCoS-Calling\n | \nLLaMA-CoS | \n100.00 | \n91.57 | \n95.56 | \n98.88 | \n94.18 | \n98.55 | \n95.61 | \n88.44 | \n
| Task | \n\n \n\n\nIf in\n\nCoS-GPT\n \n | \n\n \n\n\nCoS\n\nStage\n \n | \n\n \n\n\nLLaMA\n\n-CoS\n \n | \n\n \n\n\nChatGPT\n \n | \n
| \n\n\nAQUA-RAT\n | \n\\usym2713 | \nPlanning | \n58.80 | \n52.90 | \n
| Calling | \n56.12 | \n65.94 | \n||
| \n\n\nMATH\n | \n\\usym2713 | \nPlanning | \n65.83 | \n52.61 | \n
| Calling | \n50.75 | \n43.25 | \n||
| \n\n\nTabMWP\n | \n\\usym2713 | \nPlanning | \n90.00 | \n60.75 | \n
| Calling | \n66.00 | \n32.75 | \n||
| \n\n\nFinQA\n | \n\\usym2717 | \nPlanning | \n70.51 | \n50.15 | \n
| Calling | \n22.38 | \n32.38 | \n||
| \n\n\nGSM8K\n | \n\\usym2717 | \nPlanning | \n61.29 | \n53.83 | \n
| Calling | \n57.25 | \n36.50 | \n
| Task | \n\n \n\n\nToolkit Origin\n \n | \n\n \n\n\nLLaMA-CoS\n \n | \n\n \n\n\nChatGPT\n \n | \n
| \n\n\nDynamic\n\nCnt.\n | \nRaw | \n97.50 | \n80.83 | \n
| From Dyck L. | \n73.30 | \n79.17 | \n|
| \n\n\nUnit\n\nInterp.\n | \nRaw | \n70.83 | \n80.83 | \n
| From Arith. | \n65.83 | \n80.00 | \n