Weight-space programming solves a ladder of two-player games zero-shot

Zero-Shot Weight-Space Programming Across a Difficulty-Graded Ladder of Two-Player Games

Artificial Intelligence

Summary

This paper studies whether large language models can set the exact weights of neural networks by hand to play games, without any training or trial and error. The authors build networks that play games of increasing difficulty, from simple Nim to a chess endgame, and check their moves with independent proofs or oracles. They find that the networks work perfectly at simpler levels, but performance drops at harder chess tasks, showing limits in the approach and architecture. Their work also releases all their networks and tools for future comparison.

What this means in practice

  • For game ai developers: Create exact game-playing networks without training data or gradient updates, useful for straightforward or provably correct AI agents.
  • For neural network architects: Evaluate architectural limits of weight-fixed networks by testing zero-shot programming across varying game difficulties.

Authors

Eduardo C. Garrido-Merchán

Abstract

Weight-space programming sets every weight of a feed-forward network by hand, from an explicit specification of the desired computation, with no gradient descent and no training data anywhere in the construction. We ask how far a large language model can carry this practice, zero-shot, across games of deliberately increasing difficulty. We design a four-rung ladder (Nim, tic-tac-toe, two sizes of Connect Four, and a chess endgame) spanning four axes of difficulty, and build a second, independent network at each rung, using a structurally unrelated game at matched difficulty, to separate what a rung's finding says about the difficulty level from what it says about the one game used to measure it. Ten networks are built and independently verified. Six admit a complete, non-statistical proof of correctness against an independent oracle. The chess-endgame rung is exact where correctness is cheap, move legality and, in one network, an exact one-ply mate detector, and collapses where it is expensive, composing a multi-move mating plan, from 84.1-100% success at the one-ply tactics it was built to get right down to 0.57-9.13% in full adversarial play; its replicate shows this collapse is an order of magnitude less severe under a shorter required plan and exact rather than heuristic safety checks. The standard-sized rung confirms, by holding hand-set evaluation weights fixed and adding only shallow search, that the remaining gap is an architectural ceiling rather than a matter of engineering effort, and this exact collapse reproduces across two structurally unrelated games. We release all ten networks, their independent oracles, construction-cost measurements, and a cold-prompt protocol for a future cross-model comparison