A CPU Inside an LLM: Zero-Gradient Compilation of Exact Circuits into Pretrained Weights

Media reports Hugging Face Papers Source language: Chinese

Researchers introduced closed-form, zero-gradient circuit-grafting primitives that compile deterministic computation directly into pretrained LM weights without training examples or runtime hooks. On 6-digit arithmetic benchmarks, IBM Granite 4.2 3B and Alibaba Qwen 3.5 4B achieved 99.17% and 100.00% ALU-prefix accuracy respectively.

Full text

Pretrained language models (LMs) remain unreliable on exact multi-digit arithmetic without gradient fine-tuning or external tool interpreters. We introduce closed-form, zero-gradient circuit-grafting primitives that compile deterministic computation directly into pretrained LM weights using zero training examples and zero runtime hooks, while bounding per-layer normalization drift and preserving natural-language (NL) competence. Our construction combines three analytical mechanisms: (1) two-layer Mellin-GLU quadrature over bias-free SwiGLU/GeGLU layers, turning unknown per-token RMSNorm scaling into translation along a log-spaced gate grid (>250times dynamic range) with IEEE 754 error bounds; (2) cross-token scale cancellation via signed Chebyshev Indefinite-RoPE self-sink attention and memoryless SiLU-antisymmetric GatedDeltaNet shifts; and (3) axis-aligned width extension (d to d+K) with d/(d+K) norm compensation, Base-1000 canonical BPE emission, and differential (+254,-256) tied unembeddings. An automated 5-stage compiler synthesizes 4-8 digit signed {+,-,times} circuits in constant TC^0 depth alongside 3-literal Boolean (100%, N=1{,}728; plus 42-layer Gemma 4 E4B audits) and 5-bit Majority/Parity (100%) circuits. On 6-digit arithmetic (N=1{,}200, BF16), IBM Granite 4.2 3B (1.07M-1.33M active non-zeros, le 0.036%) and Alibaba Qwen 3.5 4B (2.25M-2.39M non-zeros, le 0.057%) reach 99.17% and 100.00% ALU-prefix accuracy (100% with an FP32 sidecar) and 98%-100% mid-sentence resumption, outperforming 12.8K-sample LoRA (42.2% / 61.7%, 0% inline continuation). A factorized block-sparse sidecar holds A100 memory at 7.89 / 10.65 GiB (4.60 / 8.14 GiB under INT8/INT4 + FP32 sidecar) while passing all 14/14 pre-registered >2K-token Tango splits (N=4{,}392; 0/11,134 OOD triggers).

View arXiv page View PDF Add to collection

Sources

← Back to Daily AI Report