Escaping the BLEU Trap: A Signal-Grounded Framework with Decoupled Semantic Guidance for EEG-to-Text Decoding

Researchers propose SemKey, a novel framework that addresses key limitations in EEG-to-text decoding by preventing hallucinations and improving semantic fidelity through decoupled guidance objectives. The system redesigns neural encoder-LLM interaction and introduces new evaluation metrics beyond BLEU scores to achieve state-of-the-art performance in brain-computer interfaces.

• →SemKey framework solves three major problems in EEG-to-text decoding: semantic bias, signal neglect, and inflated BLEU evaluation metrics.
• →The system uses four decoupled semantic objectives (sentiment, topic, length, surprisal) to ground text generation in actual neural signals.
• →Novel architecture injects semantic prompts as queries and EEG embeddings as key-value pairs to force attention on neural inputs.
• →New evaluation protocols using N-way Retrieval Accuracy and Fréchet Distance provide more rigorous assessment than traditional metrics.
• →Research demonstrates effective elimination of hallucinations and achieves state-of-the-art performance on robust evaluation protocols.