FLEXITOKENS replaces rigid subword tokenizers and fixed-compression auxiliary losses with a simplified boundary-prediction objective in byte-level models, yielding lower over-fragmentation and up to 10-point gains on multilingual and domain-adaptation tasks.
Zero-shot tokenizer transfer
2 Pith papers cite this work. Polarity classification is still indexing.
2
Pith papers citing it
verdicts
UNVERDICTED 2representative citing papers
Manticore-Deep uses tiled Bayesian field-level inference on SDSS and BOSS data to produce posterior ensembles of 3D cosmic fields that are consistent with LCDM and validated by 7.4σ CMB lensing and 3.5σ kSZ detections.
citing papers explorer
-
FLEXITOKENS: Flexible Tokenization for Evolving Language Models
FLEXITOKENS replaces rigid subword tokenizers and fixed-compression auxiliary losses with a simplified boundary-prediction objective in byte-level models, yielding lower over-fragmentation and up to 10-point gains on multilingual and domain-adaptation tasks.
-
The Manticore Project II: Bayesian digital twins of cosmic structure across the SDSS and BOSS volumes
Manticore-Deep uses tiled Bayesian field-level inference on SDSS and BOSS data to produce posterior ensembles of 3D cosmic fields that are consistent with LCDM and validated by 7.4σ CMB lensing and 3.5σ kSZ detections.