Focal Interest
Focal Interest

Inference

  1. Conditional splitting holds for c-representations and weakly consistent bases
  2. Integrated ACD models can imply infinite-mean durations
    Asymptotic theory for integrated autoregressive conditional duration models reveals infinite-mean trading intervals in cryptocurrency ETFs, requiring new inference methods.
  3. Survey maps graph roles in retrieval-augmented generation
    Survey of graph-based techniques in retrieval-augmented generation systems, examining their roles in database construction, algorithms, and reasoning with structured knowledge.
  4. Mental health symptoms predict flexible inference use
    Study shows that specific mental health symptoms predict how people use flexible reasoning when making complex inferences, with alterations in goal-directed decision-making as a key mechanism.
  5. RNN-based distortion models improved CAT bond pricing
    Catastrophe bond pricing framework combining distortion operator theory with recurrent neural networks, capturing discontinuous repricing and tail-risk compensation.
  6. Forensic knowledge depends on managing transformation, uncertainty, and warrant
    Theoretical analysis of forensic knowledge production addressing technical instrumentalism through ontology, transformation processes, and epistemic capture mechanisms in institutional structures.
  7. Maximal effort linked to greater preference for correct performance
    Active inference model reveals voluntary mental effort is governed by motivation for accuracy rather than inhibition of habitual responses in Stroop task performance.
  8. Task learning increased redundancy in macaque visual cortex responses
    Task learning increases information redundancy in macaque visual cortex while preserving individual neuron information, supporting Bayesian inference over efficiency-based models.
  9. Liger+ dynamically balances latency and throughput in large model inference
    Distributed inference system using interleaved parallelism to dynamically balance latency-throughput trade-offs via task-aware batch management and strategic kernel scheduling across multiple GPUs.
  10. Russian cognitive distortions form a hierarchical co-occurrence network
    Study validates cognitive distortion architecture in Russian discourse using AI analysis of 249,414 social media texts, identifying 443,447 distortion instances across 18 categories with stable.