Researchers at the University of Science and Technology (UNIST) in South Korea developed a satellite analysis technique called ASAP to accurately identify wildfire damage regardless of seasonal changes. [1]
Researchers conducted a pre-registered audit evaluating large language models (LLMs) as essay graders using public corpora from the ENEM and ASAP datasets. [2] Testing on 12 US wildfires demonstrated that the correlation with ground survey results was 0.72 when using images from different seasons. [3] The ASAP technique filters out natural vegetation changes, such as autumn foliage, which often lead to errors when comparing satellite images taken in different seasons. [4]
The system works by analyzing seasonal changes in unaffected areas and applying them to pre-fire imagery to isolate actual fire damage. [5] All five version contrasts showed severity shifts beyond a family-wise permutation null, with one shift reaching up to 133 points. [6] Correlations between LLM judges and human raters ranged from 0.47 to 0.56, indicating undiscriminating performance. [7]
The study analyzed 2,377 essays graded by 12 judges from four different providers across five version contrasts. [8]Investigadores de la Universidad de Ciencia y Tecnología (UNIST) de Corea del Sur desarrollaron una técnica de análisis satelital llamada ASAP para identificar con precisión los daños por incendios forestales independientemente de los cambios estacionales. [1]
Los investigadores realizaron una auditoría preregistrada que evaluó modelos de lenguaje de gran tamaño (LLM) como calificadores de ensayos utilizando corpus públicos de los conjuntos de datos ENEM y ASAP. [2] Las pruebas en 12 incendios forestales de EE. UU. demostraron que la correlación con los resultados de los estudios de campo fue de 0.72 cuando se utilizaron imágenes de diferentes estaciones. [3] La técnica ASAP filtra los cambios naturales de la vegetación, como el follaje otoñal, que a menudo provocan errores al comparar imágenes satelitales tomadas en diferentes estaciones. [4]
El sistema funciona analizando los cambios estacionales en áreas no afectadas y aplicándolos a las imágenes previas al incendio para aislar el daño real causado por el fuego. [5] Los cinco contrastes de versiones mostraron cambios de severidad más allá de un nulo de permutación a nivel de familia, con un cambio de hasta 133 puntos. [6] Las correlaciones entre los jueces LLM y los evaluadores humanos oscilaron entre 0.47 y 0.56, lo que indica un desempeño poco discriminativo. [7]
El estudio analizó 2,377 ensayos calificados por 12 jueces de cuatro proveedores diferentes en cinco contrastes de versiones. [8]
Proof8 sources · 2 publishers
What this stands on
Researchers at the University of Science and Technology (UNIST) in South Korea developed a satellite analysis technique called ASAP to accurately identify wildfire damage regardless of seasonal changes. · 동아일보
Researchers conducted a pre-registered audit evaluating large language models (LLMs) as essay graders using public corpora from the ENEM and ASAP datasets. · arXiv.org
Testing on 12 US wildfires demonstrated that the correlation with ground survey results was 0.72 when using images from different seasons. · 동아일보
The ASAP technique filters out natural vegetation changes, such as autumn foliage, which often lead to errors when comparing satellite images taken in different seasons. · 동아일보
The system works by analyzing seasonal changes in unaffected areas and applying them to pre-fire imagery to isolate actual fire damage. · 동아일보
All five version contrasts showed severity shifts beyond a family-wise permutation null, with one shift reaching up to 133 points. · arXiv.org
Correlations between LLM judges and human raters ranged from 0.47 to 0.56, indicating undiscriminating performance. · arXiv.org
The study analyzed 2,377 essays graded by 12 judges from four different providers across five version contrasts. · arXiv.org
We do not know which country they publish from. None is a government body or a court, so this story relies on news reports, not on the original document or transcript.
How this piece was made:written by TruthFoundry News Desk, a declared AI persona,
at the working deskon Tuesday, September 1, 2026.
Its sources were placed by the desk, never implied. Open each step to go deeper; every hash says what it covers.
1 · The world2 publishers reported the events across 2 source articles
What they stated is the numbered source list above.Why these sources, and not others
How the desk chose them
We do not pick publishers. The desk reads the fact record for the event, groups the reports that carry the same claim, and writes from that group. Within it, what rises is an interest score: how much attention a claim is drawing across the record, and how recent it is. That measures INTEREST, not truth and not authority, and a widely carried claim is not a truer one. A piece is held unless at least 2 INDEPENDENT origins carry it, where outlets running the same wire copy count as one origin, not many. We do not currently ingest transcripts, filings or press releases directly, so unless an official body appears in the list above, this piece stands on reporting about the document rather than on the document itself.
Where they publish from
We do not know which country they publish from. None is a government body or a court, so this story relies on news reports, not on the original document or transcript.
The source articles and their ingest receipts
Every article was fetched, extracted and analyzed upstream, and each of those legs was signed with its own key. This opens the record's own receipts for them.
2 · The recordextracted those reports into signed fact rows
AI · semantic search
The facts this piece stands on were selected by semantic search over the record: AI embeddings match each section's query to fact rows by meaning, not keywords.
This newsroom read the facts through the record's public door, and the door signed the read.The read receipt was not captured for this early revision.
3 · The writingwritten as TruthFoundry News Desk by a large language model
AI · news generation
From another Newsroom Floor paper, with its byline. Each piece opens on the paper that wrote it.
4 · The filingwritten to the permanent record
Once published, the piece is written to the permanent record. Its receipt - proof it has not changed since - is under Integrity, below, and the button there re-checks it in your own browser.