LLMDB.APP RESEARCH ARCHIVE

AI Research Paper Archive & Publications

Browse through 256 peer-reviewed AI publications, arxiv preprints, technical abstracts, and key takeaway summaries covering LLM architectures, vision models, alignment, and multi-agent systems.

RESEARCH FEED

AI Research Paper Feed

Curated index of breakthrough publications in artificial intelligence. Access synthesized TL;DRs, key technical abstracts, and direct citations.

agentsAug 6, 2026
The Bitter Lesson of Tool Calling

By Ishan Patel, Sahil Sen, Elias Lumer et al.

Programmatic tool calling (tools as Python stubs) matches or beats JSON tool calling across 14 LLMs on BFCL v4, with >10% gains for GPT-5.6 and robust parallel/context-rot performance.

alignmentAug 6, 2026
Learning When to Trust via Selective Context Preference Optimization

By Xian Sun, Wei Chow, Yingshuo Wang et al.

Introduces SC2W metric for context-induced answer flips and SCOPE, a DPO variant that optimizes balanced preference pairs across clean/misleading/correct/irrelevant contexts, reducing susceptibility without sacrificing accuracy.

otherAug 6, 2026
Investigating Artificial Intelligence Digital Sovereignty in Mobile Shopping Apps: A Case Study of Nigeria

By George Grispos, Sajda Qureshi

Static forensic analysis of Nigerian shopping apps reveals AI features are widespread but disclosure is minimal, enabling an empirical transparency index to assess individual digital sovereignty.

otherAug 6, 2026
An Optimal Agnostic PAC Algorithm

By Markus Engelund Mathiasen, Jian Qian, Nikita Zhivotovskiy

Constructs the first agnostic PAC learner that exactly matches Devroye et al.'s lower bounds for every fixed L*, using a data-dependent hypothesis selection scheme to adapt to noise levels.

agentsAug 6, 2026
Tracing the Heart: An Evidence-Linked Pipeline for Heart-Failure Feature Engineering

By Soorya Ram Shimgekar, Michelle Hu, Dorisa Shehi et al.

A multi-agent LLM pipeline automates heart-failure feature engineering from EHR tables, generating auditable, rubric-scored aggregates that lift phenotyping AUROC to 0.96 with provenance tracking.

llmAug 5, 2026
Toward Skill-Native LLMs: Skill Entropy for Benchmarking and Training Long-Horizon Reasoning

By Yinghui He, Ling Yang, Jiarui Liu et al.

Introduces Skill Entropy to quantify cross-skill switching difficulty, proposes Skill^2-Bench with 558 skills, and Skill-Entropy RL using skill-sequence alignment rewards—boosting Qwen3-4B from 34.4% to 68.4%.