Line of inquiry
Inquiring lines›Where does language-model reasonin…›How do modularity, routing, and se…›this line of inquiry
How should we design LLM systems to maintain alignment and control?
A broader line of inquiry — a family of 17 specific questions the research asks around this. Follow one into its inquiring-line page, or move sideways to a related line below.
Questions in this line of inquiry 17
Specific inquiring lines the field asks around this — ordered from the most general framing down to the most specific angle.
- How do different LLM integration paradigms affect inheritance of pretraining biases?
- How can human-centered objectives be embedded earlier in the LLM pipeline?
- What deployment feedback loops amplify LLM pretraining popularity in live systems?
- What interaction controls matter most for effective human-LLM collaboration?
- How does content-only knowledge in LLMs enable pretraining popularity to leak through?
- What unique perspective do designers bring to LLM adaptation that engineers might miss?
- How does this differ from using LLMs as the policy itself?
- How does the outer loop escape its own LLM's knowledge boundaries when discovering mechanisms?
- What biases might an LLM judge introduce into an on-policy alignment process?
- What types of tasks benefit most from dynamically generated interfaces?
- What implicit knowledge about catalogs do LLMs learn from ranking signals alone?
- What makes the embers of autoregression framework predictive?
- What role does KL penalty strength play in format selection?
- How does direct web access change privacy assumptions built on API limits?
- Can utility control modify LLM values more effectively than output filtering?
- How does KL penalty strength affect the degree of format collapse during RL?
- How do aligned LoRA adapters compose through parameter-space arithmetic?