LLMs 能成为计算机吗？

Language models can solve tough math problems at research grade but struggle on simple computational tasks that involve reasoning over many steps and long context. Even multiplying two numbers or solving small Sudokus is nearly impossible unless they rely on external tools.

語言模型可以在研究級別解決困難的數學問題，但對於涉及多步推理和長上下文的簡單計算任務卻很掙扎。即使相乘兩個數字或解決小型 Sudokus 也幾乎不可能，除非依賴外部工具。

But what does it take for an LLM itself to be as reliable and efficient as a computer?

但是，要让 LLM 本身像计算机一样可靠和高效，需要什么？

We answer this by literally building a computer inside a transformer. We turn arbitrary C code into tokens that the model itself can execute reliably for millions of steps in seconds.

我们通过在 transformer 内部字面意义上构建一台计算机来回答这个问题。我们将任意 C code 转换为令牌，模型本身可以在几秒钟内可靠地执行数百万步。

Here is how it works when solving an optimization problem that proceeds in many steps, namely min-cost perfect matching via the Hungarian algorithm.

以下是它在解决一个多步进行的优化问题时的运作方式，即通过 Hungarian algorithm 进行的 min-cost perfect matching。

Solve the min-cost perfect matching for this 10×10 cost matrix:

解决这个 10×10 成本矩阵的最小成本完美匹配：

61 58 35 86 32 39 41 27 21 42

59 77 97 99 78 21 89 72 35 63

88 85 37 57 59 97 37 29 69 94

32 82 53 20 77 96 21 70 50 61

15 44 81 10 64 36 56 78 20 69

76 35 87 69 16 55 26 37 30 66

86 32 74 94 32 14 24 12 31 70

97 63 20 64 90 21 28 49 89 10

58 52 27 76 61 35 17 91 37 66

42 79 61 26 55 98 70 17 26 86

31,621 tok/s384,118 tokens6,874 lines/s

21213720151612101712345678910

The model does not call an external tool. Instead, it executes the program directly via its transformer weights, producing an execution trace token by token and streaming results at more than 30k tokens/sec on a CPU.

模型不调用外部工具。相反，它通过其 transformer 权重直接执行程序，按 ...