Qwen3-Next models struggle to lower cognitive demand in Bloom-aligned educational control

Researchers introduce a Bloom-aligned framework to measure the ability of Large Language Models to preserve instructional intent while shifting cognitive demand toward specific learning objectives. The study evaluates Qwen3-Next-80B-A3B-Instruct and Qwen3-Coder-Next across 2,520 programming tasks from three benchmarks.

• Both models reliably increase cognitive demand but struggle to lower it, revealing a robust directional asymmetry.
• General difficulty control and Bloom's control were tested using revised Bloom's Taxonomy as the operational scale.
• The general model shows clearer middle-layer separability for both control contrasts compared to the coder model.
• Semantic-delta clustering and layer-wise Fisher's Discriminant Ratio probing characterized these outcomes.

The results indicate that strong execution performance does not automatically entail Bloom-aligned educational control.