PUBMETAGet the briefing
VOL. 01FREE & OPENEST. 2026

PubMeta

We read the
methods.

PubMeta is a free publication for hard science and technical research—reporting what was tested, what the evidence supports, and where uncertainty remains.

Explore the research How we evaluate evidence
PMPUBMETA

OUR SIGNALMethodsDataLimits

SCROLL TO INSPECTOPEN LINKS · NO PRESS-RELEASE SCIENCE
01 Featured analysis
Editor’s selection · Space

The authors combine a nominal controller, a structured residual policy, and a recursive online safety filter for coupled translation-and-attitude control. The available abstract reports better tracking and training convergence than pure policy learning and single-step shielding, with near-zero simulated safety violations—but supplies no numerical results or evidence of flight or hardware testing.

Primary paper: Yuhan Sun, Di Huang, Zhijun Li · Guidance Navigation and Control (2026)
Read the evidence brief
02 Research library

Browse the evidence.

Filter source-linked evidence briefs and landmark references.

8source-linked entries
in the library
Sources before summaries. Public evidence briefs are versioned editorial records. Landmark fallback entries link directly to their primary sources and remain clearly separate from published PubMeta analysis.
01
Space·2026·4 min brief

A safety filter reins in residual reinforcement learning for simulated spacecraft formations

Yuhan Sun, Di Huang, Zhijun Li · Guidance Navigation and Control

The authors combine a nominal controller, a structured residual policy, and a recursive online safety filter for coupled translation-and-attitude control. The available abstract reports better tracking and training convergence than pure policy learning and single-step shielding, with near-zero simulated safety violations—but supplies no numerical results or evidence of flight or hardware testing.

Published record; peer review not independently verified Publisher record and PDF endpoint present; primary license and unrestricted access not verified No public code, data, or consolidated reproducibility artifact verified

Boundary: Publisher Cloudflare controls blocked direct inspection of the landing page and PDF, leaving equations, proofs, tables, figures, numerical results, and experimental details unavailable. The number of spacecraft, orbit regime, simulator, scenarios, episodes, seeds, uncertainty distributions, comparator implementations, filter-intervention rates, runtimes, ablations, and out-of-distribution tests could not be verified. The abstract reports simulations only—no hardware-in-the-loop, air-bearing, robotic, or on-orbit experiment. Near-zero violations are not violation-free performance, and moderate computation cost does not establish compatibility with an onboard processor or control cycle. No code, data, independent replication, or consolidated artifact was located.

Study typeSimulation study of safety-aware residual reinforcement learningSample / dataSimulated leader–follower formation control; scenario counts, seeds, numerical outcomes, and hardware tests were not accessible
Read our brief Primary source
02
Computing·2026·4 min brief

A three-model pipeline generates and scores synthetic UML diagrams, but its benchmark and validation remain preliminary

Van-Viet Nguyen et al. · Discover Artificial Intelligence

The authors connect a small instruction model, a reasoning LLM, PlantUML, and three vision-language evaluators across nine diagram types. They report a 4.23 automated score and correlation with human ratings, but the corpus is synthetic, the principal comparison is only five examples, and several dataset and result counts are difficult to reconcile.

Peer-reviewed journal article Open accepted manuscript under CC BY-NC-ND 4.0 Fragmented public datasets; no consolidated DOI-bound release or executable code repository verified

Boundary: All requirements were model-generated rather than collected from real projects. Only English prompts were tested. The paper reports no external end-to-end benchmark, repeated runs, decoding parameters, random seeds, uncertainty intervals, or fully documented reasoning ablation. Human-study assignment, blinding, inter-rater reliability, and diagram-level results are not reported. The visible tables total 14,957 records rather than 15,000 and do not support a balanced corpus; failure and success counts are also difficult to reconcile. Public artifacts are fragmented, no consolidated DOI-bound release or executable pipeline repository was verified, and no independent reproduction was found.

Study typeMethods paper with synthetic benchmark and cross-sectional human-rating validationSample / dataReported 15,000 records across nine UML types; result tables total 14,957; 90 diagrams in human validation with 155 participants
Read our brief Primary source
03
AI·2026·4 min brief

Personal DNA–RNA training improves expression prediction for familiar genes, but not for unseen loci

Shumin Li, Ruibang Luo, Yuanhua Huang · Molecular Systems Biology

In GEUVADIS lymphoblastoid cells, xDecoder adapted genomic foundation-model embeddings using paired personal genomes and RNA measurements. Predictions improved across new individuals for genes represented during training, while DNA-only transfer to genes on held-out chromosomes remained near zero. Matched ATAC-seq rescued part of that failure, but an ATAC-only control showed that much of the gain came from measured chromatin state itself.

Peer-reviewed journal article Open access under CC BY 4.0 Public source data and MIT-licensed code; end-to-end reproduction package incomplete

Boundary: The primary personalized cohort was small, the data came from one immortalized lymphoblastoid-cell context, and there was no independent cohort, tissue, or prospective replication. The gene set was enriched for known European cis-eQTLs, the unseen-locus result used one fixed chromosome split, and transfer weakened in YRI individuals. Personal sequences omitted indels and structural variants. Main models used final-epoch selection without a general validation set. The code repository lacks a tagged release, a fully pinned environment, and some final-paper analysis scripts. The results are not suitable for clinical or reliable per-gene use.

Study typeComputational methods and comparative benchmarking studySample / data462 GEUVADIS individuals; primary 50-train/100-test split; 3,259 genes including 200 personalized-training genes and 287 unseen loci
Read our brief Primary source
04
AI·2021·8 min brief

Highly accurate protein structure prediction with AlphaFold

Jumper et al. · Nature

A neural-network system for predicting three-dimensional protein structures from amino-acid sequences, evaluated in the CASP14 blind assessment.

Peer reviewed Open access Code available

Boundary: Benchmark performance does not remove uncertainty for every protein, complex, ligand state, or dynamic conformation.

Study typeBenchmark / methodsSample / dataCASP14 targets
Read our brief Primary source
05
Computing·2017·6 min brief

Attention Is All You Need

Vaswani et al. · NeurIPS

Introduces the Transformer, an attention-based sequence architecture evaluated primarily on machine translation benchmarks.

Peer reviewed Open access Code available

Boundary: The original experiments are task- and scale-specific; later capabilities should not be attributed to this paper alone.

Study typeArchitecture / benchmarkSample / dataWMT translation tasks
Primary source
06
Biology·2022·7 min brief

The complete sequence of a human genome

Nurk et al. · Science

Reports a gapless assembly for almost all chromosomes in a human cell line, resolving regions absent from earlier references.

Peer reviewed Open access Data available

Boundary: A single unusual cell line is not a representation of global human genomic diversity.

Study typeGenome assemblySample / dataOne complete hydatidiform-mole genome
Primary source
07
Medicine·2015·9 min brief

Efficacy and safety of RTS,S/AS01 malaria vaccine with or without a booster dose in infants and children in Africa

RTS,S Clinical Trials Partnership · The Lancet

Final results from a multi-site trial evaluating efficacy and safety of the RTS,S/AS01 malaria vaccine over extended follow-up.

Peer reviewed Open access Protocol available

Boundary: Efficacy varied by age, dose schedule, site, and follow-up period; results require context rather than a single headline number.

Study typePhase 3 randomized trialSample / data15,459 infants and children
Primary source
08
Materials·2009·5 min brief

Organometal halide perovskites as visible-light sensitizers for photovoltaic cells

Kojima et al. · Journal of the American Chemical Society

An early demonstration of organometal halide perovskites as absorbers in photovoltaic cells, helping establish a major research direction.

Peer reviewed Open access Supporting info

Boundary: Early devices had modest efficiency and poor liquid-electrolyte stability; it is not evidence of commercial readiness.

Study typeLaboratory experimentSample / dataPrototype photovoltaic cells
Primary source
03 Fields of inquiry

One method. Many scales.

From molecular structures to planetary systems.

01

AI + COMPUTING

Models, systems, benchmarks, and the infrastructure shaping machine intelligence.

AI · Computing
02

LIFE SCIENCES

Genomes, cells, organisms, and the tools used to observe living systems.

Biology · Medicine
03

PLANET + MATTER

Climate, energy, materials, and measurements across planetary scales.

Climate · Materials
04

FRONTIERS

Observations and instruments extending what we know beyond Earth.

Space
04 The PubMeta Method

Evidence before
excitement.

Every brief uses the same inspection frame, so readers can see not only what a paper says, but why—and how confidently.

Follow the project
  1. 01

    Start at the source

    We link the primary paper, identify its publication status, and distinguish author claims from our interpretation.

  2. 02

    Inspect the design

    Study type, comparison, sample or data, measurement, and analysis come before conclusions.

  3. 03

    Map the uncertainty

    Limitations, missing evidence, conflicts, and the boundary of generalization stay visible.

  4. 04

    Keep a public record

    Sources, corrections, metadata verification, and meaningful updates remain attached to the brief.