Research

Published work from AIME Research on measuring cognitive coherence in artificial and biological minds.

Level of Consciousness Signatures Across Biological and Artificial Minds

Jamaludheen K N · AIME Research · 2026

A unified framework for measuring cognition in human EEG and large language models. We present four studies validating that the same 13-function cognitive model applies to both biological and artificial networks. Seven of 13 functions show same-direction effects across substrates. Causal validation via layer isolation achieves 6.5× chance detection.

Studies

  1. Token-level cognitive function detection in LLMs (12/13 functions significant at p < 0.001)
  2. Sentence-level classification across 5 model architectures
  3. Phase transition detection at cognitive state boundaries
  4. Within-sentence zero-confound validation
Read Full Paper →Plain-Language Summary →

5-Model Cognitive Benchmark

AIME Research · 2026

The first published benchmark of cognitive coherence in open-source LLMs. True Coherence scores for Llama-3.3-70B (15.37%), Mistral-Small-24B (11.46%), Qwen3.5-35B-A3B (10.24%), Qwen3.5-Distilled (9.07%), and Gemma-3-12B (7.44%).

View Benchmark →

Cross-Substrate EEG Validation

AIME Research · 2026

Validation of the LOC framework on 21 EEG subjects performing cognitive tasks. Classification accuracy 2.31× above chance. Seven of 13 cognitive functions show convergent patterns between human brains and language models.