Dimensions overall score 7.0

GitHub Code Pulse

No public code linked for this paper yet.

• Existing image captioning methods for Early Childhood Education lack large-scale, domain-specific datasets, limiting their ability to capture fine-grained semantic concepts unique to ECE scenarios.limitation 90%
• Conventional training paradigms for image captioning exhibit limitations in enhancing professional object description capability.limitation 85%
• The ECAC benchmark comprises 256,121 real-world images annotated with expert-level captions and fine-grained labels for ECE daily activity image captioning.method 95%
• ECAC includes a domain-oriented evaluation protocol called Teaching Toy Recognition Score (TTS) to explicitly measure professional object naming accuracy.method 90%
• The proposed RSRS training framework dynamically alternates between reinforcement learning and supervised fine-tuning by rerouting hard samples with zero rewards to supervised optimization.method 90%
• RSRS effectively mitigates advantage collapse and enables stable optimization for fine-grained recognition in image captioning.technical 85%
• The KinderMM-Cap-3B model achieves a TTS score of 51.06, substantially outperforming state-of-the-art baselines.result 95%
• The proposed approach maintains superior caption quality while achieving high TTS scores, highlighting its potential for specialized educational applications.result 80%

Competitive landscape

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

Startup potential card