Dimensions overall score 7.0

GitHub Code Pulse

No public code linked for this paper yet.

• Standard LLM benchmarks only evaluate the assistant turn and leave unmeasured whether the model encodes awareness of what follows the assistant response.limitation 95%
• User-turn generation serves as a probe to measure interaction awareness in language models.method 90%
• Interaction awareness in LLMs is decoupled from task accuracy.result 85%
• Within the Qwen3.5 family, GSM8K accuracy scales from 41% to 96.8% across model sizes, yet genuine follow-up rates under deterministic generation remain near zero.result 90%
• Higher temperature sampling reveals interaction awareness is latent with follow-up rates reaching 22%.result 85%
• Controlled perturbations validate that the proposed probe measures a real property of the model.method 80%
• Collaboration-oriented post-training on Qwen3.5-2B demonstrates an increase in follow-up rates.result 80%
• User-turn generation captures a dimension of LLM behavior (interaction awareness) that is unexplored and invisible with current assistant-only benchmarks.result 90%

Competitive landscape

Competitor map is still being generated for this paper. Enable generation or check back soon.

Startup potential card