CORTEXA
← Browse
arxivcs.AIcs.CL2026-07-09

Different Teachers, Different Capabilities: Sub-1B On-Device Distillation for Structured Text Enrichment

Vinay Kumar Chaganti

High-volume structured extraction pays a large model's latency on every item, so distilling the task into a small on-device model is attractive: comparable output at a fraction of the time and cost. We measure what that distillation actually delivers, per sub-task. Each news article is mapped to one JSON object with a short summary and five categorical labels. We distill an 8B reasoning teacher (deepseek-r1:8b) into a 0.6B student (Qwen3-0.6B; QLoRA, three seeds), and add two teacher controls: a same-size non-reasoning teacher and a larger managed pipeline. A blinded, reference-free, three-judge panel scores every arm against the full article, alongside two non-distillation baselines, few-shot prompting and constrained decoding. The student runs at about 0.8 s per article against the teacher's 39 s, and recovers 58% of the base-to-teacher gap on summary quality, beating its primary baseline (constrained decoding) by +16.8 points and few-shot prompting by a secondary +4.9. A same-size non-reasoning teacher trains a student no better than the untuned base, so the summary gain follows from the teacher's reasoning nature rather than its scale. Capabilities then split by teacher: the reasoning teacher transfers writing quality and the managed pipeline transfers label diversity, while a same-size instruction teacher's students stay more grounded on the 22 short, thin-source articles in the 93-item test set (74 versus 55 faithful), where the reasoning-lineage student fabricates. That grounding difference is a consistent ordering rather than a significant aggregate effect, and the subgroup is small, so we report it as a direction. Because no single engine wins every field, the deliverable is a per-field routing map for on-device enrichment.

View free PDFSource page

Related papers

arxivcs.CLcs.AI2026-07-24

From Isolated Tasks to Structured Capabilities: A Multilayer Taxonomy for Large Language Models

Shixin Fang, Jiachen Wo, Wenjuan Qin, Sihang Jiang, Yanghua Xiao

Large language model (LLM) evaluation spans diverse tasks and benchmarks, yet evidence remains organized around tasks rather than the capabilities they probe. This fragmentation limits cross-study comparison, obscures capabilities tasks recruit, and makes coverage gaps difficult…

View free PDFSource page
arxivcs.CLcs.AI2026-06-29

ARKD: Adaptive Reinforcement Learning-Guided Bidirectional KL Divergence Distillation for Text Generation

Zilong Liu, Xuewen Zhang, Jinrui Xing, Juyi Qiao, Huiyong Wang, Junming Jiao

Knowledge distillation (KD) is a key technique for compressing Large Language Models (LLMs), yet methods relying on a single KL objective often fail to balance primary distribution fitting with long-tail probability modeling, limiting both generation quality and generalization. T…

View free PDFSource page
arxivcs.CLcs.AIcs.PF2026-07-01

BaseRT: Best-in-Class LLM Inference on Apple Silicon via Native Metal

Prabod Rathnayaka, Fabian Waschkowski, Lukas Wesemann

We present BaseRT, a native Metal inference runtime for large language models (LLMs) on Apple Silicon, and report the highest inference throughput on this hardware to date. Existing runtimes, including llama.cpp and MLX-based frameworks, incur overhead from abstractions not desig…

View free PDFSource page