Show HN: Duplicate 3 layers in a 24B LLM, logical deduction .22→.76. No training

I replicated Ng's RYS method and found that duplicating 3 specific layers in Qwen2.5-32B boosts reasoning by 17% and duplicating layers 12-14 in Devstral-24B improves logical deduction from 0.22→0.76 on BBH — no training, no weight changes, just routing hidden states through the same circuit twice. Tools included. Two AMD GPUs, one evening.

Duplicate 3 layers. No training. Logical deduction goes from 0.22 → 0.76.

This toolkit finds and exploits "reasoning circuits" hidden inside transformer models. The idea: certain contiguous blocks of layers act as indivisible cognitive units. Duplicate them in the forward pass — same weights, no training, no merging — and the model gets measurably smarter on specific capabilities.

Built on David Ng's RYS method and extended with new findings. Everything here was discovered on two AMD consumer GPUs (RX 7900 XT + RX 6950 XT) in one evening.

Validated on standard benchmarks via lm-evaluation-harness at n=50:

| Benchmark | Base | +3 layers | Change | |---|---|---|---| | BBH Logical Deduction | 0.22 | 0.76 | +245% | | GSM8K (strict) | 0.48 | 0.64 | +33% | | MBPP (code gen) | 0.72 | 0.78 | +8% | | GSM8K (flexible) | 0.82 | 0.86 | +5% | | BBH Navigate | 0.96 | 0.98 | +2% | | BBH Date Understanding | 0.82 | 0.84 | +2% | | BBH Causal Judgement | 0.66 | 0.66 | — | | IFEval (strict) | 0.68 | 0.68 | — |

Average improvement: +8% across all metrics. Nothing degraded.

Measured on custom probe suite (BBH-derived + EQ-Bench-style + GSM8K):

| Probe | Base | +3 layers | Change | |---|---|---|---| | Reasoning (causal + logic + nav) | 76.5% | 94.1% | +23% | | EQ (emotional intelligence) | 92.1 | 93.6 | +1.6% |

Transformers organize themselves during training into functional circuits — multi-layer processing units that perform complete cognitive operations. These circuits are indivisible: duplicating a single layer does almost nothing, but duplicating the right block of 3-4 layers gives the model a second pass through its reasoning pipeline.

Different models have different circuits in different places:

Devstral-24B(40 layers): reasoning circuit at layers12-14****Qwen2.5-32B(64 layers): reasoning circuit at layers7-9

The boundaries are sharp. Shift the block by one layer in either direction and the improvement disappears or inverts.

Different duplication patterns create distinct cognitive profiles from the same weights:

| Pattern | Math | EQ | Character | |---|---|---|---| | Double-pass 13-16 | ↑↑ | ↑ | Math specialist | | Triple-pass 13-16 | ↑ | ↑↑ | EQ specialist | | Interleaved 13,13,14,14,15,15,16 | ↑↑↑ | ↓ | Pure math mode | | Quadruple-pass 13-16 | — | ↑↑ | EQ mode, math neutral |

Same weights on disk. Same VRAM for the base model. Just different routing.