PromptCBLUE gives a practical, Chinese-language testbed for medical LLM products. It shows that inexpensive PEFT fine-tuning of open 13B models can beat few-shot API use, so companies can invest in targeted fine-tuning to improve medical features.

Co^2PT offers a low-cost way to reduce downstream gender bias: it freezes the main model, tunes small prompts, and avoids costly full-model retraining while lowering fairness gaps on real tasks.

If you run face recognition on mobile or edge devices, distilling a high-performing Transformer into a CNN can boost verification accuracy substantially while keeping hardware-friendly inference.

You can cut LLM layers to dramatically shrink model size and lower hosting and fine-tuning costs while keeping or improving classification accuracy on many few-shot tasks.

Prompt tuning lets teams deploy clinical summarization with much lower compute and faster turnaround than full fine-tuning while often improving quality if you have a large domain LLM.

Tuning only soft prompts on a frozen billion‑scale clinical LLM cuts compute and deployment costs while keeping or improving cross‑site and few‑shot extraction accuracy.

RadioLLM lets you reuse LLM priors for multiple radio tasks, improving classification and denoising while cutting prompt overhead and latency in many benchmark scenarios.

BOOST improves generation commonsense without fine-tuning large models, so teams can upgrade deployed LMs cheaply by adding a small controller and a scorer.

You can steer deployed LLMs at inference without costly human labels or weight updates; train a small prefix once to get near-zero runtime cost and plug it into existing models to enforce safety or tone constraints.

SAMPLE helps detect multimodal fake news with far fewer labels and far fewer trainable parameters than full fine-tuning, lowering data and compute costs for real-world monitoring systems.

Convert multi-call agent pipelines into a single finetuned model with stage tokens to reduce API calls, cut token costs, and produce more structurally coherent section drafts for academic-style content.

LoRAP lets teams deploy low-bit GNNs with much of the original accuracy retained while keeping memory and speed gains; it is a small, trainable add-on compatible with existing QAT pipelines.

One universal encoding plus small scenario prompts reduces model sprawl, lowers serving cost via KV-cache reuse, and improves live business metrics, so businesses can swap many specialized models for a single adaptable pipeline.

SVDecode can raise model accuracy or truthfulness by a few percentage points with no extra inference memory and little engineering effort, speeding deployment and reducing compute for task adaptation.

Switching to a single soft-prompted model cuts model count and maintenance while keeping similar accuracy; using LLM-assisted label transfer lowers non-English annotation costs and speeds roll-out across languages.

Optimizing prompts often improves model outputs without costly retraining; however, inconsistent evaluations hide how well methods generalize, so businesses should validate prompt methods on their own balanced data before production.