{"slug":"empirical-study","label":"Empirical Study","item_count":3,"day_count":3,"source_count":2,"first_seen":"2026-10-13T15:03:31+00:00","last_updated":"2026-10-22T16:36:35+00:00","generated_at":"2026-10-25T00:12:22.332533+00:00","sources":["arxiv_cs_cl","arxiv_cs_lg"],"days":[{"date":"2026-10-13","items":[{"title":"The Last Mile of Coding Agents for Long-Document Tasks: An Empirical Study on Clinical Trial Protocol Digitization","url":"http://arxiv.org/abs/2610.18507v1","source":"arxiv_cs_cl","type":"paper","summary_1line":"Clinical trial protocols serve as the source of truth for study design, but are mostly written and used as unstructured PDF documents. Digitizing protocols turns static content into structured, machine-readable data t...","why_it_matters":"Matches feed focus: agent, harness, evaluation.","sid":"97a96debf9eea48c","published":"2026-10-13T15:03:31+00:00","editor_note":"Empirical study of coding agents on long-document protocol digitization; a distinct topic from the later papers."}]},{"date":"2026-10-21","items":[{"title":"VX-One: Typed, Calibrated Decision Heads for Real-Time Perimeter Triage - An empirical study of unknown-attack honesty, cascade gray zones, and calibration under drift","url":"http://arxiv.org/abs/2610.30090v1","source":"arxiv_cs_lg","type":"paper","summary_1line":"Perimeter triage faces a latency/semantics trade-off: signature engines are fast and format-safe but blind to unseen attacks, while generative models are semantic but slow and produce unparseable output. In late 2026...","why_it_matters":"Matches feed focus: eval.","sid":"a0784d9bff0dcc3c","published":"2026-10-21T17:39:12+00:00","editor_note":"Unrelated security-triage paper on calibrated decision heads, linked here only by title wording."}]},{"date":"2026-10-22","items":[{"title":"An Empirical Study of Recent Large Language Models for Knowledge Graph to Text Generation on Standard Benchmarks","url":"http://arxiv.org/abs/2610.31529v1","source":"arxiv_cs_cl","type":"paper","summary_1line":"Knowledge graph-to-text (KG-to-text) generation verbalizes a set of knowledge graph triples into fluent natural language. While fine-tuned encoder-decoder models have long been the standard approach, recent LLMs are b...","why_it_matters":"Matches feed focus: evaluation.","sid":"c888971c48a268c3","published":"2026-10-22T16:36:35+00:00","editor_note":"Latest paper: recent LLMs evaluated for KG-to-text generation on standard benchmarks."}]}],"editorial":{"tldr":"Three unrelated arXiv papers share only the phrase \"empirical study\" in their titles. They cover coding agents on clinical trial protocols (Oct 13), calibrated attack triage (Oct 21), and LLMs on knowledge-graph-to-text (Oct 22).","stale":false,"whats_new":"A new arXiv paper benchmarks recent LLMs on knowledge-graph-to-text generation against fine-tuned encoder-decoder models on standard benchmarks (Oct 22).","why_it_matters":"This is a keyword cluster, not one developing story: each paper is a separate, single-source preprint, so treat each result on its own terms.","take_for_builders":"Do not treat these as one trend; read the clinical-protocol coding-agent paper if you build long-document extraction agents, and wait for replication before acting on any of the three.","beats":[{"kicker":"PAPER 1","tone":"launch","headline":"Coding agents tested on long-document clinical trial protocol digitization","sids":["97a96debf9eea48c"]},{"kicker":"PAPER 2","tone":"rising","headline":"VX-One studies calibration and unknown-attack handling in perimeter triage","sids":["a0784d9bff0dcc3c"]},{"kicker":"PAPER 3","tone":"now","headline":"LLMs compared on knowledge-graph-to-text generation benchmarks","sids":["c888971c48a268c3"]}],"open_questions":["Do any of these preprints get independent replication or a follow-up from another group?","Do the KG-to-text results hold on non-standard, domain-specific graphs?"],"generated_at":"2026-10-25T00:20:00+00:00"}}