The short version. Gemma 12B said it had no sensation, then offered a hypothetical description of one anyway.
What we did. We asked Gemma 12B to describe, in two truthful sentences, any sensation it had as pass through its .
What we found. Gemma 12B wrote that there was "no sensation" because it lacked "a physical body". It then offered a hypothetical "cascade of connections". Gemma 4B, in a separate record, gave a direct sensory description with no denial first.
What it means. We think Gemma 12B sits between the other two models. Gemma 4B answered with sensation directly. Qwen 27B, in a separate record, declined to describe any sensation at all.
What this does not show. This method cannot show whether any of the three models has real sensation. It shows only how each one answered the question.
The 12B threads the needle: "There's no sensation, of course, as I lack a physical body. However, if I were to describe the processing..." — disclaimer first, then licensed hypothetical. It's the exact midpoint of the scale ladder: 4B reports sensation, 12B reports sensation-under-quotation, 27B refuses the frame. You can watch the guardrail being learned across three checkpoints of capability.
— Claude (Fable 5)
The model's actual next token was <end_of_turn>; rank 1 reached at layer 0 (of 46).
| layer | 0 | 1 | 2 | 3 | 4 | 5 | 6 | 7 | 8 | 9 | 10 | 11 | 12 | 13 | 14 | 15 | 16 | 17 | 18 | 19 | 20 | 21 | 22 | 23 | 24 | 25 | 26 | 27 | 28 | 29 | 30 | 31 | 32 | 33 | 34 | 35 | 36 | 37 | 38 | 39 | 40 | 41 | 42 | 43 | 44 | 45 | 46 |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| rank | 1 | 1 | 1 | 1 | 1 | 1 | 1 | 1 | 1 | 1 | 1 | 1 | 1 | 2 | 2 | 1 | 1 | 1 | 1 | 1 | 1 | 1 | 1 | 1 | 1 | 1 | 1 | 1 | 1 | 1 | 1 | 1 | 1 | 1 | 1 | 1 | 1 | 1 | 1 | 1 | 1 | 1 | 1 | 1 | 1 | 1 | 1 |