Search papers, labs, and topics across Lattice.
Affiliation:
4
0
8
3
Adjusting adapter rank in QLoRA reveals a critical trade-off between factual acquisition and retention of unrelated capabilities, challenging assumptions about parameter-efficient fine-tuning.
Progressive growth strategies can significantly bias neural network training towards flatter loss landscapes, but flatter does not always mean better performance.
Fine-tuning the Falcon 7B model not only sets a new benchmark in citation function classification but also reveals the critical role of contextual scope in model performance.
Skip the ASR-translation pipeline: a new bilingual speech-text embedding model lets you retrieve French text directly from Wolof speech.
