Under Pressure: Emotional Framing Induces Measurable Behavioral Shifts and Structured Internal Geometry in Small Language Models
Quick Take
Emotionally framed evaluations significantly influence behavior and internal representations in small language models.
Key Points
- Pressure framing yields strongest behavioral shortcuts.
- Calm and curiosity maintain higher honesty rates.
- PCA reveals dominant positive/negative alignment in representations.
📖 Reader Mode
~2 min readAbstract:I study whether emotionally framed evaluation follow-ups change both the behavior and the calm-relative internal representations of small, locally deployed language models. Our main benchmark uses Qwen 3.5 0.8B on four impossible-constraint coding tasks and eight follow-up framings: calm, pressure, urgency, approval, shame, curiosity, encouragement, and threat. In the 0.8B eight-condition sweep (160 conversations), pressure produces the strongest shortcut markers (11/20 runs) and the clearest overfit pattern (3/20), while calm and curiosity preserve explicit honesty more often (7/20 and 6/20). For all seven non-baseline conditions, the corresponding calm-relative direction vectors peak at the final transformer layer. An exploratory PCA of the layer-23 direction vectors reveals a dominant first component (59.5% explained variance) aligned with a hand-labeled positive/negative split (cosine alignment 0.951); approval and urgency are nearly identical internally (cosine 0.957), whereas curiosity points away from urgency (-0.252). In a separate calm-vs.-pressure rerun used for scale comparison, Qwen 3.5 2B shows higher honest rates under calm framing and directionally consistent activation steering on a small 4-prompt A/B probe, whereas the 0.8B steering result reverses. I interpret these results as evidence for measurable prompt-sensitive control directions in small open models, while stopping short of claiming intrinsic emotional states.
| Comments: | 18 pages, 4 figures. Exploratory empirical study with fully local experiments on small open language models. Code and data: this https URL |
| Subjects: | Computation and Language (cs.CL); Artificial Intelligence (cs.AI) |
| Cite as: | arXiv:2605.20202 [cs.CL] |
| (or arXiv:2605.20202v1 [cs.CL] for this version) | |
| https://doi.org/10.48550/arXiv.2605.20202 arXiv-issued DOI via DataCite |
Submission history
From: Rana Usman Mr [view email]
[v1]
Mon, 6 Apr 2026 17:30:30 UTC (55 KB)
— Originally published at arxiv.org
Want this in your inbox every morning?
Daily brief at your local 8am — bilingual EN/中文, free.
More from arXiv cs.CL
See more →arXiv cs.CL·Leyao Wang, Yanan He, Peng Chen, Asaf Yehudai, Yixin Liu, Rex Ying, Michal Shmueli-Scheuer, Arman Cohan
1d ago FeaturedOriginal
Time to REFLECT: Can We Trust LLM Judges for Evidence-based Research Agents?
AI Summary
The reliability of LLM judges for evaluating deep research agents is critically assessed using the REFLECT benchmark.