Video understanding AI company Filer said on Oct. 21 that two of its research papers were accepted at the AI conference Conference on Neural Information Processing Systems (NeurIPS) 2026.

The accepted papers are Continuous Audio Thought (CoAT), which enables AI to understand nonverbal information in speech such as emotion and intonation, and Geometric Anchor Preference Optimization (GAPO), which helps AI learn stably even from data that contain errors. The two papers are scheduled to be presented on Dec. 10 in Sydney, Australia.

CoAT is a technology that directly reflects information such as emotion, intonation and pitch—often lost when converting speech to text—into the AI inference process. When applied to three voice AI models, 38 of 42 performance evaluation metrics improved.

In particular, the speech emotion recognition accuracy of the Qwen2.5-Omni model increased to about 79 percent from about 55 percent.

GAPO is a technology that reduces the impact of incorrect or uncertain data in AI training processes that use human preference evaluations. It measures the stability of evaluation results by making small changes to the model being trained, and reduces the training weight of unstable data.

Filer applied the technology to four AI models and compared performance with eight existing training methods. It said it confirmed overall strong performance in response quality and user preference evaluations, and showed stable performance even in an experiment that inserted errors into 20 percent of the training data.

Filer plans to use the research results for Brand Safety, which manages advertising so it is not exposed to inappropriate content, and for content review technology.

Jongseok Kim (김종석), Filer's head of AI, said that when determining the safety of a video, important factors can include not only the scenes on screen but also the tone of voice and background sounds. He said the company will continue to develop related technology so it can analyze such information without missing it.