{
  "source": "SABER_AAAI2026_CR.pdf, p. 5, Tables 1\u20132",
  "metricNote": "accuracy in percent, length is average thinking tokens",
  "1.5B": {
    "Base": {
      "MATH500": {
        "accuracy": 78.9,
        "thinkTokens": 8042
      },
      "GSM8K": {
        "accuracy": 82.1,
        "thinkTokens": 3471
      },
      "MATH": {
        "accuracy": 80.6,
        "thinkTokens": 7866
      },
      "MBPP": {
        "accuracy": 38.9,
        "thinkTokens": 5755
      }
    },
    "DeepThink": {
      "MATH500": {
        "accuracy": 83.2,
        "thinkTokens": 5353
      },
      "GSM8K": {
        "accuracy": 85.7,
        "thinkTokens": 1947
      },
      "MATH": {
        "accuracy": 85.2,
        "thinkTokens": 4748
      },
      "MBPP": {
        "accuracy": 53.7,
        "thinkTokens": 3010
      }
    },
    "CoreThink": {
      "MATH500": {
        "accuracy": 82.1,
        "thinkTokens": 3294
      },
      "GSM8K": {
        "accuracy": 83.1,
        "thinkTokens": 930
      },
      "MATH": {
        "accuracy": 84.3,
        "thinkTokens": 3045
      },
      "MBPP": {
        "accuracy": 49.8,
        "thinkTokens": 1254
      }
    },
    "FastThink": {
      "MATH500": {
        "accuracy": 81.4,
        "thinkTokens": 2899
      },
      "GSM8K": {
        "accuracy": 82.5,
        "thinkTokens": 778
      },
      "MATH": {
        "accuracy": 83.5,
        "thinkTokens": 2719
      },
      "MBPP": {
        "accuracy": 45.1,
        "thinkTokens": 942
      }
    },
    "Base-NoThink": {
      "MATH500": {
        "accuracy": 65.1,
        "thinkTokens": 0
      },
      "GSM8K": {
        "accuracy": 61.1,
        "thinkTokens": 0
      },
      "MATH": {
        "accuracy": 65.5,
        "thinkTokens": 0
      },
      "MBPP": {
        "accuracy": 30.7,
        "thinkTokens": 0
      }
    },
    "NoThink": {
      "MATH500": {
        "accuracy": 76.3,
        "thinkTokens": 0
      },
      "GSM8K": {
        "accuracy": 78.1,
        "thinkTokens": 0
      },
      "MATH": {
        "accuracy": 76.9,
        "thinkTokens": 0
      },
      "MBPP": {
        "accuracy": 44.7,
        "thinkTokens": 0
      }
    }
  },
  "7B": {
    "Base": {
      "MATH500": {
        "accuracy": 91.5,
        "thinkTokens": 8221
      },
      "GSM8K": {
        "accuracy": 91.4,
        "thinkTokens": 3402
      },
      "MATH": {
        "accuracy": 92.4,
        "thinkTokens": 7838
      },
      "MBPP": {
        "accuracy": 59.9,
        "thinkTokens": 5021
      },
      "LiveBench-R": {
        "accuracy": 36.4,
        "thinkTokens": 5619
      }
    },
    "DeepThink": {
      "MATH500": {
        "accuracy": 91.9,
        "thinkTokens": 6265
      },
      "GSM8K": {
        "accuracy": 92.0,
        "thinkTokens": 2351
      },
      "MATH": {
        "accuracy": 92.9,
        "thinkTokens": 5763
      },
      "MBPP": {
        "accuracy": 65.8,
        "thinkTokens": 2426
      },
      "LiveBench-R": {
        "accuracy": 38.3,
        "thinkTokens": 4401
      }
    },
    "CoreThink": {
      "MATH500": {
        "accuracy": 90.1,
        "thinkTokens": 2820
      },
      "GSM8K": {
        "accuracy": 91.7,
        "thinkTokens": 1319
      },
      "MATH": {
        "accuracy": 91.2,
        "thinkTokens": 2557
      },
      "MBPP": {
        "accuracy": 63.0,
        "thinkTokens": 1556
      },
      "LiveBench-R": {
        "accuracy": 32.1,
        "thinkTokens": 3039
      }
    },
    "FastThink": {
      "MATH500": {
        "accuracy": 86.5,
        "thinkTokens": 1563
      },
      "GSM8K": {
        "accuracy": 90.3,
        "thinkTokens": 495
      },
      "MATH": {
        "accuracy": 87.2,
        "thinkTokens": 1478
      },
      "MBPP": {
        "accuracy": 61.1,
        "thinkTokens": 969
      },
      "LiveBench-R": {
        "accuracy": 30.6,
        "thinkTokens": 2166
      }
    },
    "Base-NoThink": {
      "MATH500": {
        "accuracy": 79.1,
        "thinkTokens": 0
      },
      "GSM8K": {
        "accuracy": 86.4,
        "thinkTokens": 0
      },
      "MATH": {
        "accuracy": 79.7,
        "thinkTokens": 0
      },
      "MBPP": {
        "accuracy": 43.2,
        "thinkTokens": 0
      },
      "LiveBench-R": {
        "accuracy": 17.9,
        "thinkTokens": 0
      }
    },
    "NoThink": {
      "MATH500": {
        "accuracy": 85.1,
        "thinkTokens": 0
      },
      "GSM8K": {
        "accuracy": 89.2,
        "thinkTokens": 0
      },
      "MATH": {
        "accuracy": 85.5,
        "thinkTokens": 0
      },
      "MBPP": {
        "accuracy": 58.8,
        "thinkTokens": 0
      },
      "LiveBench-R": {
        "accuracy": 26.8,
        "thinkTokens": 0
      }
    }
  }
}