Theme of inquiry
How are AI-generated and human-written texts perceived differently?
A question within its area, explored through 6 lines of inquiry below — each a family of specific questions the research asks.
43 specific questions
- At what point does output quality outweigh diversity value in synthetic data tasks?
- Why does preference tuning reduce diversity in code but increase it in creative tasks?
- When does RLHF reduce diversity and when does it preserve semantic variation?
- How does diversity collapse during iterative self-improvement cycles?
- How do quality, diversity, and complexity create different effects on downstream model performance?
- What makes external diversity more effective than sequential revision steps?
- What conditions make training diversity better than individual expert quality?
31 specific questions
- When does the right constraint beat additional model capacity?
- How should tiny language models be architected differently than large ones?
- Do small models show different parameter efficiency patterns than large models?
- Can a two-layer network outgeneralize billion-parameter models through recursion alone?
- Why does the right structural prior matter more than raw model capacity?
- What role does inductive bias play versus model capacity in practice?
- Can width-scaling replace depth-scaling on inherently sequential problems?
19 specific questions
- Can mid-tier models benefit more from self-generated harness updates than others?
- Can harness updates benefit agents equally across all model sizes?
- What components of agent scaffolding most impact domain-specific output quality?
- What cognitive burdens should move from model parameters into harness infrastructure?
- Can harness evolution be redirected toward distilling transferable procedures instead?
- Why do mid-tier models benefit more from memorized harness shortcuts?
- Why do evolved harness edits mostly memorize rather than generalize?
17 specific questions
- Does architectural discovery follow an empirical scaling law like neural networks?
- What scaling laws govern autonomous architecture discovery in AI systems?
- Do autonomous architecture discoveries follow predictable scaling laws like human research?
- Can the scaling law for discovery extend beyond architectures to agentic systems?
- Why do scaling laws fail to predict optimal architectures at small parameter counts?
- How do conditional scaling laws incorporate hardware into architecture choices?
- Why do human-designed neural architectures eventually get replaced by learned ones?
13 specific questions
- Does specialized training in one domain create capability cliffs elsewhere?
- How does over-specialization create capability cliffs outside target domains?
- What causes models to develop domain capability cliffs after specialization?
- Why does over-specialization create a domain capability cliff in LLMs?
- Why do fine-tuned models fail outside their specialized domains?
- What capability risks emerge when models are optimized for single domains?
- What distinguishes domain-specific failure modes from general model limitations?
39 specific questions
- Can identical model performance mask fundamentally broken internal representations?
- How do weight perturbations reveal what performance benchmarks cannot measure?
- How do surface statistical regularities enable correct outputs while degrading robustness?
- Can RL format selection explain performance gains attributed to algorithmic improvements?
- Why do internal representations differ when external performance matches?
- Does highlighting input features reduce human over-reliance on machine outputs?
- Why do only two of fourteen models improve when problem constraints are removed?