Inclusion-of-Thoughts: Mitigating Preference Instability via Purifying the Decision Space

Build an 'IoT-wrapper' for inference APIs that filters distractors in MCQs before the LLM processes them. This will increase robustness for automated exam-taking agents.

Suggested repo: think-filter

"Make your LLM immune to distracting multiple-choice noise."

Estimated effort: 20h