📑 PageIndex: Document Index for Vectorless, Reasoning-based RAG
agentic-ai
agents
ai
ai-agents
context-engineering
information-retrieval
llm
rag
reasoning
retrieval
retrieval-augmented-generation
vector-database
Updated 2026-04-23 18:00:08 +02:00
I replicated Ng's RYS method and found that duplicating 3 specific layers in Qwen2.5-32B boosts reasoning by 17% and duplicating layers 12-14 in Devstral-24B improves logical deduction from 0.22→0.76 on BBH — no training, no weight changes, just routing hidden states through the same circuit twice. Tools included. Two AMD GPUs, one evening.
Updated 2026-03-20 02:51:23 +01:00