Selected work

Representative Papers

Proper Body Landmark Subset Enables More Accurate and 5X Faster Recognition of Isolated Signs in LIBRAS

Oct 28, 2025

Lightweight keypoint detection for Brazilian Sign Language (LIBRAS) isolated-word recognition faces a fundamental trade-off between recognition accuracy and inference speed. Method: This paper proposes an efficient recognition framework based on optimized keypoint subset selection and missing-keypoint compensation. It employs MediaPipe for body keypoint extraction, introduces a discriminability-driven strategy to identify the minimal effective keypoint subset for sign motion representation, and applies spline interpolation to recover keypoints lost due to occlusion or detection failure. Contribution/Results: The method preserves skeletal motion fidelity while significantly reducing computational redundancy. Experiments demonstrate that, compared to state-of-the-art approaches, our method achieves a 1.2% absolute accuracy gain, a 5.3× speedup in inference latency, and a 68% reduction in model parameters—enabling scalable, robust sign language recognition under resource-constrained conditions.

0 citationsRead paper

Echoes of Power: Investigating Geopolitical Bias in US and China Large Language Models

Mar 20, 2025

This study investigates ideological and cultural biases embedded in large language models (LLMs) regarding geopolitics, specifically comparing ChatGPT (U.S.-developed) and DeepSeek (China-developed). Method: Leveraging a curated geopolitical question set, we conduct a multimodal evaluation—integrating qualitative discourse analysis with quantitative metrics including stance polarity, factual density, and rhetorical bias—to assess cross-model consistency on sensitive political topics. Contribution/Results: Our analysis reveals significant yet non-binary divergence: the models exhibit high response agreement on 42% of questions, challenging the assumption of technological determinism in geopolitical alignment. Crucially, they demonstrate substantial convergence on foundational factual claims, indicating latent capacity for cross-ideological factual alignment. This work provides the first empirical, multidimensional assessment of mainstream U.S. and Chinese LLMs on geopolitical discourse, offering both methodological innovation—via integrated qualitative-quantitative evaluation—and empirical grounding for AI value embedding research and transnational LLM governance frameworks.

0 citationsRead paper

Recent publications

Latest Papers

Proper Body Landmark Subset Enables More Accurate and 5X Faster Recognition of Isolated Signs in LIBRAS

Oct 28, 2025

Lightweight keypoint detection for Brazilian Sign Language (LIBRAS) isolated-word recognition faces a fundamental trade-off between recognition accuracy and inference speed. Method: This paper proposes an efficient recognition framework based on optimized keypoint subset selection and missing-keypoint compensation. It employs MediaPipe for body keypoint extraction, introduces a discriminability-driven strategy to identify the minimal effective keypoint subset for sign motion representation, and applies spline interpolation to recover keypoints lost due to occlusion or detection failure. Contribution/Results: The method preserves skeletal motion fidelity while significantly reducing computational redundancy. Experiments demonstrate that, compared to state-of-the-art approaches, our method achieves a 1.2% absolute accuracy gain, a 5.3× speedup in inference latency, and a 68% reduction in model parameters—enabling scalable, robust sign language recognition under resource-constrained conditions.

0 citationsRead paper

Echoes of Power: Investigating Geopolitical Bias in US and China Large Language Models

Mar 20, 2025

This study investigates ideological and cultural biases embedded in large language models (LLMs) regarding geopolitics, specifically comparing ChatGPT (U.S.-developed) and DeepSeek (China-developed). Method: Leveraging a curated geopolitical question set, we conduct a multimodal evaluation—integrating qualitative discourse analysis with quantitative metrics including stance polarity, factual density, and rhetorical bias—to assess cross-model consistency on sensitive political topics. Contribution/Results: Our analysis reveals significant yet non-binary divergence: the models exhibit high response agreement on 42% of questions, challenging the assumption of technological determinism in geopolitical alignment. Crucially, they demonstrate substantial convergence on foundational factual claims, indicating latent capacity for cross-ideological factual alignment. This work provides the first empirical, multidimensional assessment of mainstream U.S. and Chinese LLMs on geopolitical discourse, offering both methodological innovation—via integrated qualitative-quantitative evaluation—and empirical grounding for AI value embedding research and transnational LLM governance frameworks.

0 citationsRead paper