Topic / Trend Rising

Domain-Specific AI and Scientific Tools

Open models and tools for legal retrieval, healthcare, education, data science and photonic design indicate a shift toward specialized, deployable AI systems.

Detected: 2026-09-02 · Updated: 2026-09-02

Related Coverage

2026-09-01 • ArXiv cs.AI

Clinical LLMs: the real bottleneck is not knowledge, it's choosing between guideline pathways

Nine frontier LLMs evaluated on 2,005 oncology decision points show a collective boundary: in 42.1% of items none answers correctly, especially when choosing between guideline pathways. Two decisiveness-tuned models make unsafe commitments 3–5 times ...

#Hardware #LLM On-Premise #Fine-Tuning

2026-09-01 • ArXiv cs.AI

DS-Lighting: Making the Agent Harness Explicit for Data-Science Automation

DS-Lighting is an open-source toolkit that makes the agent harness explicit for data-science automation. It decomposes the harness into four reusable layers—data, workflow, execution, and evaluation—and represents agents as executable operator progra...

#Hardware #LLM On-Premise #DevOps

2026-08-31 • Microsoft Research

GigaPath-Flash and GigaTIME-Flash: efficiency for population-scale pathology

Microsoft Research, the University of Washington, and Providence have released two open-weight pathology models. GigaPath-Flash distills a billion-parameter encoder into a 22M backbone, reducing compute by about 50 times while retaining 97% of perfor...

#Hardware #LLM On-Premise #Fine-Tuning

2026-08-28 • ArXiv cs.AI

PICasso and the Value of Physical Constraints: LLMs Alone Are Not Enough in Photonic Design

PICasso combines natural-language specifications, GDS generation, and physical verification to design photonic integrated circuits. On a benchmark of 36 parameterized tasks, the framework reaches up to 92.7% structural satisfaction and reduces mean i...

#LLM On-Premise #DevOps

2026-08-28 • ArXiv cs.AI

EduRiskX: a neuro-symbolic route to early academic risk prediction

A neuro-symbolic framework combines a Transformer predictor with F-Logic rules to identify at-risk students in online education. On OULAD it reaches accuracy 0.900 and F1-score 0.894, with average detection at week 9.32. The key difference is explain...

#LLM On-Premise #DevOps

2026-08-27 • ArXiv cs.LG

GreenLeaf Law Embed Tiny and the Local Turn in Legal Retrieval

A 0.6 billion parameter legal embedding model shows that distillation, hard negative mining, and binary quantization can move legal retrieval toward self-hosted deployments. The structural issue is not benchmark performance alone, but the ability to ...

2026-08-27 • ArXiv cs.LG

GreenLeaf Law Embed Tiny: A Compact Embedding Model for Legal Retrieval

GreenLeaf Law Embed Tiny is a 0.6B parameter embedding model for legal retrieval. It achieves 75.11% on MLEB and 64.38% on MTEB Law. Its two-stage pipeline combines distillation and fine-tuning with hard negative mining on 3.4 million query-passage p...

#LLM On-Premise #Fine-Tuning #DevOps

← Back to All Topics