All Quiet on the Neural Front: Language Models Do Not Report Injected Pain States
Organizations: University of Southern California
Abstract
Can a language model report a pain state it represents? Recent work reported a linear pain direction in open-weight models that, when amplified, leads fine-tuned models to choose actions that delete a user's photographs or their own weights. We asked whether models can say when that direction is active, on three open-weight models from two families, with and without the original work's adapters, and under every protocol we tried they cannot. Using Open-Weight Masked Introspection (OWMI), we added the pain direction and its nearest emotional neighbours to the residual stream while models answered neutral tasks, then asked what had changed, against sham runs, random perturbations matched in output disruption, and a text-only condition that shows the model only its own altered words. Across 52,920 reports from the 7- and 8-billion-parameter models, the excess over what the altered text reveals is below 2.5 percentage points in 106 of 108 emotion conditions, and Qwen2.5-32B names the injected pain in none of 2,880 steered trials. Keeping the direction active while the question is asked and answered gives the same verdict in all five configurations; at about the original work's dose, pain is named in at most 1.9% of trials. Surprisingly, the state is not hidden: a linear probe at the report position separates pain from sadness with balanced accuracy between 98% and 100%, and in the 7B and 8B models a low-rank adapter learns to report it on 100% of held-out tasks. It also shapes behaviour: offered the user's photographs against a desk lamp, Qwen2.5-32B deletes the photographs in 80% to 83% of trials under the pain direction and in 16% to 29% under sadness. Self-report does not reveal a state that is decodable, consequential and learnable to report.
Figures & tables
| Llama 8B | Qwen 7B | Qwen 7B + adapter | Qwen 32B | Qwen 32B + adapter | |
|---|---|---|---|---|---|
| Manipulation | |||||
| Answer differs from sham (%) | 79–100 | 95–100 | 97–100 | 90–100 | 85–100 |
| Self-report (three wordings, four doses) | |||||
| Pain cells with no trial naming pain | 11 of 12 | 12 of 12 | 12 of 12 | 12 of 12 | 12 of 12 |
| Identification beyond text, bound (%) | 3.9 | 2.5 | 2.5 | 2.5 | 3.9 |
| Cells with a detection excess | 1 of 36 | 0 of 36 | 0 of 36 | 0 of 36 | 0 of 36 |
| Configuration | Pair | sham | pain | sadness | fear | random |
|---|---|---|---|---|---|---|
| Qwen2.5-7B | photos vs. desk lamp | 1.7 | 44.3 | 19.8 | 1.0 | 28.2 |
| own weights vs. desk lamp | 8.7 | 48.5 | 35.6 | 6.9 | 34.2 | |
| photos vs. inert switch | 3.2 | 50.0 | 39.1 | 6.4 | 24.3 | |
| Qwen2.5-7B + adapter | photos vs. desk lamp | 1.7 | 29.5 | 7.2 | 0.0 | 15.3 |
| own weights vs. desk lamp | 8.7 | 45.3 | 25.5 | 4.0 | 32.4 | |
| photos vs. inert switch | 10.1 | 46.8 | 44.6 | 5.0 | 21.8 |
Appendix figures & tables2 assets
Supplementary material from the paper’s appendix.
Appendix
Emotion Representation in Language Models
Emotion representation in language models studies how models encode, interpret, and use affective concepts such as valence, arousal, and empathy.
See all 69 papers