Line of inquiry
Inquiring lines›How do training data and architect…›What training signals and data sel…›this line of inquiry
Can scaling alone produce compositional generalization in neural networks?
A broader line of inquiry — a family of 47 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 47
Specific inquiring lines the field asks around this — ordered from the most general framing down to the most specific angle.
- Does scaling data automatically produce compositional reasoning or just better feature encoding?
- Where do neural networks still fail at compositional generalization despite scaling?
- Does scaling model size solve compositional generalization problems?
- Can recursion alone drive generalization better than model scale?
- Can scaling alone create compositional generalization without explicit binding mechanisms?
- Why do long-context language models struggle with compositional reasoning tasks?
- How does scaling and training data enable compositional behavior without symbolic mechanisms?
- What role does query-level exposure play in enabling compositional generalization?
- Does compositional generalization emerge suddenly or improve smoothly with scale?
- Can single-hop knowledge automatically compose into multi-hop capability?
- Do larger models develop more abstract features than smaller ones?
- Why does compositional reasoning fail to explain cross-domain transfer?
- Can scaling data alone solve performance gaps on long-tail concepts?
- Why does scaling data and model size improve compositional generalization?
- How does tokenization toward corpus mean affect downstream output diversity?
- Why does the right structural prior matter more than raw model capacity?
- Why should scaling laws be understood as properties of data distribution rather than training in general?
- Why does comparison reasoning generalize better than composition reasoning?
- Why do multimodal models fail on rare and underrepresented concepts?
- Can granular function calling tasks learn composition from graph-sampled data?
- How does inductive reasoning from partial evidence enable hypothesis formation?
- How does evidence retrieval affect compositional reasoning in language models?
- What architectural alternatives can capture compositional structure beyond pooled cosine?
- Why do vision and language have different optimal scaling curves?
- How does co-occurrence statistics alone produce hierarchical concept organization?
- How do Bayesian models share statistical strength across sparse user datasets?
- Can world models form from aggregated partial information across training distributions?
- Why do frequent words rank higher in taxonomic abstraction hierarchies?
- What makes recursive structure different from other forms of compositional generalization?
- How do parameter scaling and latent vectors interact in language models?
- How do latents at the same hierarchy level become more correlated than tokens?
- What scaling laws govern the compute efficiency of latent prediction versus token prediction?
- What role does recursive generalization play in semantic drift during learning?
- Can token probability distributions extend swarm composition across different model architectures?
- What makes hierarchical reasoning effective for taxonomy induction?
- Do distributed relational tasks consistently underperform local classification across NLP domains?
- Why does token redundancy and poor readability emerge at trillion-parameter scale?
- What limits the extrapolation of learned operations like rotation and reflection?
- Does parameter composition work when adapter alignment is imperfect?
- Why do text-to-image models fail at composing multiple concepts together?
- What sampling strategies prevent nonsensical combinations when composing taxonomy nodes?
- What are the scaling law differences between vision and language learning?
- Why does projecting lattice vectors to a subfield increase shared unit distances?
- What power-law scaling patterns emerge when consistency models are trained at scale?
- How much does multi-token prediction help in protein design specifically?
- How does upward distillation transfer knowledge from smaller to larger networks?
- Do substitute networks converge differently than complement networks?