Source

Frozen but Not Always Accessible: A Representation Analysis of Genomic Language Models

Evidence concern: excluded from comparisons

Table 1 prints the same frozen value and spread (.5252+-.0262) for NT on both 'Enhancer' and 'Enhancer Types', while the fine-tuned references differ (.548 and .424). One of the two frozen cells is likely a transcription error in the source; both are kept as printed and should not support comparison until the authors clarify.Frozen but Not Always Accessible: A Representation Analysis of Genomic Language Models · Table 1, column 'NT', rows 'Enhancer' and 'Enhancer Types'

Table 1 and Table 4 disagree for HyenaDNA Splice Acceptor: Table 1 prints .4752±.0108 (recovery 50), while Table 4 prints 0.7308 ± 0.0033 (validation-selected) and 0.7360 ± 0.0052 (diagnostic grid-best). Five other HyenaDNA cells of Table 1 equal Table 4's grid-best values, but Splice Donor (.7247±.0087) equals the validation-selected value instead. Table 3 and Table 4 also name different grid-best readouts for HyenaDNA Splice All (MLP+Mean against MLP+Attn). Table 1 values are kept as printed.Frozen but Not Always Accessible: A Representation Analysis of Genomic Language Models · Table 1, column 'HyenaDNA', rows 'Splice Acceptor' and 'Splice Donor'; Table 4; Table 3, row 'Splice All'

arXiv:2608.05329v1, submitted 2026-08-05 · Preprint

Original source

What this source contributed

Records in this release that cite this source. A record can cite several sources.

2 models
8 configurations
2 protocols
1 dataset
144 evaluations
144 results
1 evidence claim

Results

Values as the source printed them. Compare results only under compatible protocols, datasets and metrics.

Value	Metric	Model or configuration	Dataset	Evaluation
.5314 unitless	MCC (NT benchmark Enhancer, frozen-embedding probe)	GENERATOR-v2-1.2B, frozen embeddings with a lightweight probe (Datta et al. 2026)	Nucleotide Transformer benchmark task data (Datta et al. 2026)	GENERATOR-v2-1.2B on NT benchmark Enhancer (frozen probe)
.580 unitless	MCC (NT benchmark Enhancer, reference fine-tuned score)	GENERATOR-v2-1.2B, fine-tuned reference score (Datta et al. 2026)	Nucleotide Transformer benchmark task data (Datta et al. 2026)	GENERATOR-v2-1.2B on NT benchmark Enhancer (reference fine-tuned)
.4305 unitless	MCC (NT benchmark Enhancer Types, frozen-embedding probe)	GENERATOR-v2-1.2B, frozen embeddings with a lightweight probe (Datta et al. 2026)	Nucleotide Transformer benchmark task data (Datta et al. 2026)	GENERATOR-v2-1.2B on NT benchmark Enhancer Types (frozen probe)
.477 unitless	MCC (NT benchmark Enhancer Types, reference fine-tuned score)	GENERATOR-v2-1.2B, fine-tuned reference score (Datta et al. 2026)	Nucleotide Transformer benchmark task data (Datta et al. 2026)	GENERATOR-v2-1.2B on NT benchmark Enhancer Types (reference fine-tuned)
.7381 unitless	MCC (NT benchmark H3, frozen-embedding probe)	GENERATOR-v2-1.2B, frozen embeddings with a lightweight probe (Datta et al. 2026)	Nucleotide Transformer benchmark task data (Datta et al. 2026)	GENERATOR-v2-1.2B on NT benchmark H3 (frozen probe)
.806 unitless	MCC (NT benchmark H3, reference fine-tuned score)	GENERATOR-v2-1.2B, fine-tuned reference score (Datta et al. 2026)	Nucleotide Transformer benchmark task data (Datta et al. 2026)	GENERATOR-v2-1.2B on NT benchmark H3 (reference fine-tuned)
.4997 unitless	MCC (NT benchmark H3K4me1, frozen-embedding probe)	GENERATOR-v2-1.2B, frozen embeddings with a lightweight probe (Datta et al. 2026)	Nucleotide Transformer benchmark task data (Datta et al. 2026)	GENERATOR-v2-1.2B on NT benchmark H3K4me1 (frozen probe)
.553 unitless	MCC (NT benchmark H3K4me1, reference fine-tuned score)	GENERATOR-v2-1.2B, fine-tuned reference score (Datta et al. 2026)	Nucleotide Transformer benchmark task data (Datta et al. 2026)	GENERATOR-v2-1.2B on NT benchmark H3K4me1 (reference fine-tuned)
.3159 unitless	MCC (NT benchmark H3K4me2, frozen-embedding probe)	GENERATOR-v2-1.2B, frozen embeddings with a lightweight probe (Datta et al. 2026)	Nucleotide Transformer benchmark task data (Datta et al. 2026)	GENERATOR-v2-1.2B on NT benchmark H3K4me2 (frozen probe)
.424 unitless	MCC (NT benchmark H3K4me2, reference fine-tuned score)	GENERATOR-v2-1.2B, fine-tuned reference score (Datta et al. 2026)	Nucleotide Transformer benchmark task data (Datta et al. 2026)	GENERATOR-v2-1.2B on NT benchmark H3K4me2 (reference fine-tuned)
.3822 unitless	MCC (NT benchmark H3K4me3, frozen-embedding probe)	GENERATOR-v2-1.2B, frozen embeddings with a lightweight probe (Datta et al. 2026)	Nucleotide Transformer benchmark task data (Datta et al. 2026)	GENERATOR-v2-1.2B on NT benchmark H3K4me3 (frozen probe)
.512 unitless	MCC (NT benchmark H3K4me3, reference fine-tuned score)	GENERATOR-v2-1.2B, fine-tuned reference score (Datta et al. 2026)	Nucleotide Transformer benchmark task data (Datta et al. 2026)	GENERATOR-v2-1.2B on NT benchmark H3K4me3 (reference fine-tuned)
.6098 unitless	MCC (NT benchmark H3K9ac, frozen-embedding probe)	GENERATOR-v2-1.2B, frozen embeddings with a lightweight probe (Datta et al. 2026)	Nucleotide Transformer benchmark task data (Datta et al. 2026)	GENERATOR-v2-1.2B on NT benchmark H3K9ac (frozen probe)
.612 unitless	MCC (NT benchmark H3K9ac, reference fine-tuned score)	GENERATOR-v2-1.2B, fine-tuned reference score (Datta et al. 2026)	Nucleotide Transformer benchmark task data (Datta et al. 2026)	GENERATOR-v2-1.2B on NT benchmark H3K9ac (reference fine-tuned)
.5036 unitless	MCC (NT benchmark H3K14ac, frozen-embedding probe)	GENERATOR-v2-1.2B, frozen embeddings with a lightweight probe (Datta et al. 2026)	Nucleotide Transformer benchmark task data (Datta et al. 2026)	GENERATOR-v2-1.2B on NT benchmark H3K14ac (frozen probe)
.605 unitless	MCC (NT benchmark H3K14ac, reference fine-tuned score)	GENERATOR-v2-1.2B, fine-tuned reference score (Datta et al. 2026)	Nucleotide Transformer benchmark task data (Datta et al. 2026)	GENERATOR-v2-1.2B on NT benchmark H3K14ac (reference fine-tuned)
.5667 unitless	MCC (NT benchmark H3K36me3, frozen-embedding probe)	GENERATOR-v2-1.2B, frozen embeddings with a lightweight probe (Datta et al. 2026)	Nucleotide Transformer benchmark task data (Datta et al. 2026)	GENERATOR-v2-1.2B on NT benchmark H3K36me3 (frozen probe)
.657 unitless	MCC (NT benchmark H3K36me3, reference fine-tuned score)	GENERATOR-v2-1.2B, fine-tuned reference score (Datta et al. 2026)	Nucleotide Transformer benchmark task data (Datta et al. 2026)	GENERATOR-v2-1.2B on NT benchmark H3K36me3 (reference fine-tuned)
.5621 unitless	MCC (NT benchmark H3K79me3, frozen-embedding probe)	GENERATOR-v2-1.2B, frozen embeddings with a lightweight probe (Datta et al. 2026)	Nucleotide Transformer benchmark task data (Datta et al. 2026)	GENERATOR-v2-1.2B on NT benchmark H3K79me3 (frozen probe)
.670 unitless	MCC (NT benchmark H3K79me3, reference fine-tuned score)	GENERATOR-v2-1.2B, fine-tuned reference score (Datta et al. 2026)	Nucleotide Transformer benchmark task data (Datta et al. 2026)	GENERATOR-v2-1.2B on NT benchmark H3K79me3 (reference fine-tuned)
.7928 unitless	MCC (NT benchmark H4, frozen-embedding probe)	GENERATOR-v2-1.2B, frozen embeddings with a lightweight probe (Datta et al. 2026)	Nucleotide Transformer benchmark task data (Datta et al. 2026)	GENERATOR-v2-1.2B on NT benchmark H4 (frozen probe)
.815 unitless	MCC (NT benchmark H4, reference fine-tuned score)	GENERATOR-v2-1.2B, fine-tuned reference score (Datta et al. 2026)	Nucleotide Transformer benchmark task data (Datta et al. 2026)	GENERATOR-v2-1.2B on NT benchmark H4 (reference fine-tuned)
.4439 unitless	MCC (NT benchmark H4ac, frozen-embedding probe)	GENERATOR-v2-1.2B, frozen embeddings with a lightweight probe (Datta et al. 2026)	Nucleotide Transformer benchmark task data (Datta et al. 2026)	GENERATOR-v2-1.2B on NT benchmark H4ac (frozen probe)
.592 unitless	MCC (NT benchmark H4ac, reference fine-tuned score)	GENERATOR-v2-1.2B, fine-tuned reference score (Datta et al. 2026)	Nucleotide Transformer benchmark task data (Datta et al. 2026)	GENERATOR-v2-1.2B on NT benchmark H4ac (reference fine-tuned)
.9630 fraction	F1 (NT benchmark Promoter All, frozen-embedding probe)	GENERATOR-v2-1.2B, frozen embeddings with a lightweight probe (Datta et al. 2026)	Nucleotide Transformer benchmark task data (Datta et al. 2026)	GENERATOR-v2-1.2B on NT benchmark Promoter All (frozen probe)
.962 fraction	F1 (NT benchmark Promoter All, reference fine-tuned score)	GENERATOR-v2-1.2B, fine-tuned reference score (Datta et al. 2026)	Nucleotide Transformer benchmark task data (Datta et al. 2026)	GENERATOR-v2-1.2B on NT benchmark Promoter All (reference fine-tuned)
.9662 fraction	F1 (NT benchmark Promoter Non-TATA, frozen-embedding probe)	GENERATOR-v2-1.2B, frozen embeddings with a lightweight probe (Datta et al. 2026)	Nucleotide Transformer benchmark task data (Datta et al. 2026)	GENERATOR-v2-1.2B on NT benchmark Promoter Non-TATA (frozen probe)
.962 fraction	F1 (NT benchmark Promoter Non-TATA, reference fine-tuned score)	GENERATOR-v2-1.2B, fine-tuned reference score (Datta et al. 2026)	Nucleotide Transformer benchmark task data (Datta et al. 2026)	GENERATOR-v2-1.2B on NT benchmark Promoter Non-TATA (reference fine-tuned)
.9525 fraction	F1 (NT benchmark Promoter TATA, frozen-embedding probe)	GENERATOR-v2-1.2B, frozen embeddings with a lightweight probe (Datta et al. 2026)	Nucleotide Transformer benchmark task data (Datta et al. 2026)	GENERATOR-v2-1.2B on NT benchmark Promoter TATA (frozen probe)
.948 fraction	F1 (NT benchmark Promoter TATA, reference fine-tuned score)	GENERATOR-v2-1.2B, fine-tuned reference score (Datta et al. 2026)	Nucleotide Transformer benchmark task data (Datta et al. 2026)	GENERATOR-v2-1.2B on NT benchmark Promoter TATA (reference fine-tuned)
.9089 fraction	F1 (NT benchmark Splice Acceptor, frozen-embedding probe)	GENERATOR-v2-1.2B, frozen embeddings with a lightweight probe (Datta et al. 2026)	Nucleotide Transformer benchmark task data (Datta et al. 2026)	GENERATOR-v2-1.2B on NT benchmark Splice Acceptor (frozen probe)
.981 fraction	F1 (NT benchmark Splice Acceptor, reference fine-tuned score)	GENERATOR-v2-1.2B, fine-tuned reference score (Datta et al. 2026)	Nucleotide Transformer benchmark task data (Datta et al. 2026)	GENERATOR-v2-1.2B on NT benchmark Splice Acceptor (reference fine-tuned)
.7754 fraction	F1 (NT benchmark Splice All, frozen-embedding probe)	GENERATOR-v2-1.2B, frozen embeddings with a lightweight probe (Datta et al. 2026)	Nucleotide Transformer benchmark task data (Datta et al. 2026)	GENERATOR-v2-1.2B on NT benchmark Splice All (frozen probe)
.978 fraction	F1 (NT benchmark Splice All, reference fine-tuned score)	GENERATOR-v2-1.2B, fine-tuned reference score (Datta et al. 2026)	Nucleotide Transformer benchmark task data (Datta et al. 2026)	GENERATOR-v2-1.2B on NT benchmark Splice All (reference fine-tuned)
.9094 fraction	F1 (NT benchmark Splice Donor, frozen-embedding probe)	GENERATOR-v2-1.2B, frozen embeddings with a lightweight probe (Datta et al. 2026)	Nucleotide Transformer benchmark task data (Datta et al. 2026)	GENERATOR-v2-1.2B on NT benchmark Splice Donor (frozen probe)
.978 fraction	F1 (NT benchmark Splice Donor, reference fine-tuned score)	GENERATOR-v2-1.2B, fine-tuned reference score (Datta et al. 2026)	Nucleotide Transformer benchmark task data (Datta et al. 2026)	GENERATOR-v2-1.2B on NT benchmark Splice Donor (reference fine-tuned)
.5243 unitless	MCC (NT benchmark Enhancer, frozen-embedding probe)	HyenaDNA-436K, frozen embeddings with a lightweight probe (Datta et al. 2026)	Nucleotide Transformer benchmark task data (Datta et al. 2026)	HyenaDNA-436K on NT benchmark Enhancer (frozen probe)
.517 unitless	MCC (NT benchmark Enhancer, reference fine-tuned score)	HyenaDNA-436K, fine-tuned reference score (Datta et al. 2026)	Nucleotide Transformer benchmark task data (Datta et al. 2026)	HyenaDNA-436K on NT benchmark Enhancer (reference fine-tuned)
.3710 unitless	MCC (NT benchmark Enhancer Types, frozen-embedding probe)	HyenaDNA-436K, frozen embeddings with a lightweight probe (Datta et al. 2026)	Nucleotide Transformer benchmark task data (Datta et al. 2026)	HyenaDNA-436K on NT benchmark Enhancer Types (frozen probe)
.386 unitless	MCC (NT benchmark Enhancer Types, reference fine-tuned score)	HyenaDNA-436K, fine-tuned reference score (Datta et al. 2026)	Nucleotide Transformer benchmark task data (Datta et al. 2026)	HyenaDNA-436K on NT benchmark Enhancer Types (reference fine-tuned)
.6926 unitless	MCC (NT benchmark H3, frozen-embedding probe)	HyenaDNA-436K, frozen embeddings with a lightweight probe (Datta et al. 2026)	Nucleotide Transformer benchmark task data (Datta et al. 2026)	HyenaDNA-436K on NT benchmark H3 (frozen probe)
.779 unitless	MCC (NT benchmark H3, reference fine-tuned score)	HyenaDNA-436K, fine-tuned reference score (Datta et al. 2026)	Nucleotide Transformer benchmark task data (Datta et al. 2026)	HyenaDNA-436K on NT benchmark H3 (reference fine-tuned)
.3315 unitless	MCC (NT benchmark H3K4me1, frozen-embedding probe)	HyenaDNA-436K, frozen embeddings with a lightweight probe (Datta et al. 2026)	Nucleotide Transformer benchmark task data (Datta et al. 2026)	HyenaDNA-436K on NT benchmark H3K4me1 (frozen probe)
.512 unitless	MCC (NT benchmark H3K4me1, reference fine-tuned score)	HyenaDNA-436K, fine-tuned reference score (Datta et al. 2026)	Nucleotide Transformer benchmark task data (Datta et al. 2026)	HyenaDNA-436K on NT benchmark H3K4me1 (reference fine-tuned)
.2792 unitless	MCC (NT benchmark H3K4me2, frozen-embedding probe)	HyenaDNA-436K, frozen embeddings with a lightweight probe (Datta et al. 2026)	Nucleotide Transformer benchmark task data (Datta et al. 2026)	HyenaDNA-436K on NT benchmark H3K4me2 (frozen probe)
.455 unitless	MCC (NT benchmark H3K4me2, reference fine-tuned score)	HyenaDNA-436K, fine-tuned reference score (Datta et al. 2026)	Nucleotide Transformer benchmark task data (Datta et al. 2026)	HyenaDNA-436K on NT benchmark H3K4me2 (reference fine-tuned)
.2215 unitless	MCC (NT benchmark H3K4me3, frozen-embedding probe)	HyenaDNA-436K, frozen embeddings with a lightweight probe (Datta et al. 2026)	Nucleotide Transformer benchmark task data (Datta et al. 2026)	HyenaDNA-436K on NT benchmark H3K4me3 (frozen probe)
.549 unitless	MCC (NT benchmark H3K4me3, reference fine-tuned score)	HyenaDNA-436K, fine-tuned reference score (Datta et al. 2026)	Nucleotide Transformer benchmark task data (Datta et al. 2026)	HyenaDNA-436K on NT benchmark H3K4me3 (reference fine-tuned)
.4299 unitless	MCC (NT benchmark H3K9ac, frozen-embedding probe)	HyenaDNA-436K, frozen embeddings with a lightweight probe (Datta et al. 2026)	Nucleotide Transformer benchmark task data (Datta et al. 2026)	HyenaDNA-436K on NT benchmark H3K9ac (frozen probe)
.581 unitless	MCC (NT benchmark H3K9ac, reference fine-tuned score)	HyenaDNA-436K, fine-tuned reference score (Datta et al. 2026)	Nucleotide Transformer benchmark task data (Datta et al. 2026)	HyenaDNA-436K on NT benchmark H3K9ac (reference fine-tuned)

Claims

Statements this source supports, each with where in the source it was checked.

recovery_definition: gfm2-20261011-datta2026-protocol-frozen-probe: Recovery definition Recovery = (frozen probe score / fine tuned score) x 100 (Equation 7). The per-cell recovery percentages and the promoter and splice average recovery rows of Table 1 are not extracted as results; they are a ratio of the two extracted scores. (Section 3.2, Equation 7; Table 1 recovery positions and the 'Promoter avg. recovery' and 'Splice avg. recovery' rows)

Models, benchmarks and other records

Models (2)

Configurations (8)

Protocols (2)

Datasets (1)

Nucleotide Transformer benchmark task data (Datta et al. 2026)

Evaluations (144)

Evidence

Source checking verifies the cited claim or transcription. It does not establish independent reproduction.

Evidence table

Inspect claims, sources and review details

Trace each statement to its source and review. A context-only reference supports the record generally; it does not verify an individual field. Source checking does not reproduce an experiment.

One row per statement and cited source. Multiple citations are not independent evaluations. Shared locators are labelled explicitly.

25 evidence rows matching the loaded filters

Claims, original sources and review scope · Release 2026-10-22-1b7cf027e4f6
Property and statement	Original source and location	Review and provenance
attributes.artifact_sha256 df36e4772562329550c924004a546812c8d76149a1b5238e4d0b0cfa8a82bb5c Source metadata	Frozen but Not Always Accessible: A Representation Analysis of Genomic Language Models Original source ↗ No field-specific location recorded Version: arXiv:2608.05329v1, submitted 2026-08-05 Retrieved: 2026-10-11T00:51:27Z	catalogued No individual claim review recorded Source has a recorded evidence concern. Consult its source page before using the claim. Audit details Field: `attributes.artifact_sha256` Source artifact SHA-256: `df36e4772562329550c924004a546812c8d76149a1b5238e4d0b0cfa8a82bb5c` Hash scope: Hash scope not separately documented; inspect source record Inspected artifact
Artifact url https://arxiv.org/pdf/2608.05329v1 Source metadata	Frozen but Not Always Accessible: A Representation Analysis of Genomic Language Models Original source ↗ No field-specific location recorded Version: arXiv:2608.05329v1, submitted 2026-08-05 Retrieved: 2026-10-11T00:51:27Z	catalogued No individual claim review recorded Source has a recorded evidence concern. Consult its source page before using the claim. Audit details Field: `attributes.artifact_url` Source artifact SHA-256: `df36e4772562329550c924004a546812c8d76149a1b5238e4d0b0cfa8a82bb5c` Hash scope: Hash scope not separately documented; inspect source record Inspected artifact
Evidence concerns 2 values source id gfm2-20261011-datta2026-source message Table 1 prints the same frozen value and spread (.5252+-.0262) for NT on both 'Enhancer' and 'Enhancer Types', while the fine-tuned references differ (.548 and .424). One of the two frozen cells is likely a transcription error in the source; both are kept as printed and should not support comparison until the authors clarify. source locator Table 1, column 'NT', rows 'Enhancer' and 'Enhancer Types' artifact sha256 df36e4772562329550c924004a546812c8d76149a1b5238e4d0b0cfa8a82bb5c review date 2026-10-11 source id gfm2-20261011-datta2026-source message Table 1 and Table 4 disagree for HyenaDNA Splice Acceptor: Table 1 prints .4752±.0108 (recovery 50), while Table 4 prints 0.7308 ± 0.0033 (validation-selected) and 0.7360 ± 0.0052 (diagnostic grid-best). Five other HyenaDNA cells of Table 1 equal Table 4's grid-best values, but Splice Donor (.7247±.0087) equals the validation-selected value instead. Table 3 and Table 4 also name different grid-best readouts for HyenaDNA Splice All (MLP+Mean against MLP+Attn). Table 1 values are kept as printed. source locator Table 1, column 'HyenaDNA', rows 'Splice Acceptor' and 'Splice Donor'; Table 4; Table 3, row 'Splice All' artifact sha256 df36e4772562329550c924004a546812c8d76149a1b5238e4d0b0cfa8a82bb5c review date 2026-10-11 Source metadata	Frozen but Not Always Accessible: A Representation Analysis of Genomic Language Models Original source ↗ No field-specific location recorded Version: arXiv:2608.05329v1, submitted 2026-08-05 Retrieved: 2026-10-11T00:51:27Z	catalogued No individual claim review recorded Source has a recorded evidence concern. Consult its source page before using the claim. Audit details Field: `attributes.evidence_concerns` Source artifact SHA-256: `df36e4772562329550c924004a546812c8d76149a1b5238e4d0b0cfa8a82bb5c` Hash scope: Hash scope not separately documented; inspect source record Inspected artifact
Licence arXiv.org perpetual non-exclusive license 1.0 (no redistribution of the bytes) Source metadata	Frozen but Not Always Accessible: A Representation Analysis of Genomic Language Models Original source ↗ No field-specific location recorded Version: arXiv:2608.05329v1, submitted 2026-08-05 Retrieved: 2026-10-11T00:51:27Z	catalogued No individual claim review recorded Source has a recorded evidence concern. Consult its source page before using the claim. Audit details Field: `attributes.licence` Source artifact SHA-256: `df36e4772562329550c924004a546812c8d76149a1b5238e4d0b0cfa8a82bb5c` Hash scope: Hash scope not separately documented; inspect source record Inspected artifact
Media type application/pdf Source metadata	Frozen but Not Always Accessible: A Representation Analysis of Genomic Language Models Original source ↗ No field-specific location recorded Version: arXiv:2608.05329v1, submitted 2026-08-05 Retrieved: 2026-10-11T00:51:27Z	catalogued No individual claim review recorded Source has a recorded evidence concern. Consult its source page before using the claim. Audit details Field: `attributes.media_type` Source artifact SHA-256: `df36e4772562329550c924004a546812c8d76149a1b5238e4d0b0cfa8a82bb5c` Hash scope: Hash scope not separately documented; inspect source record Inspected artifact
Publication status preprint Source metadata	Frozen but Not Always Accessible: A Representation Analysis of Genomic Language Models Original source ↗ No field-specific location recorded Version: arXiv:2608.05329v1, submitted 2026-08-05 Retrieved: 2026-10-11T00:51:27Z	catalogued No individual claim review recorded Source has a recorded evidence concern. Consult its source page before using the claim. Audit details Field: `attributes.publication_status` Source artifact SHA-256: `df36e4772562329550c924004a546812c8d76149a1b5238e4d0b0cfa8a82bb5c` Hash scope: Hash scope not separately documented; inspect source record Inspected artifact
Retrieved at 2026-10-11T00:51:27Z Source metadata	Frozen but Not Always Accessible: A Representation Analysis of Genomic Language Models Original source ↗ No field-specific location recorded Version: arXiv:2608.05329v1, submitted 2026-08-05 Retrieved: 2026-10-11T00:51:27Z	catalogued No individual claim review recorded Source has a recorded evidence concern. Consult its source page before using the claim. Audit details Field: `attributes.retrieved_at` Source artifact SHA-256: `df36e4772562329550c924004a546812c8d76149a1b5238e4d0b0cfa8a82bb5c` Hash scope: Hash scope not separately documented; inspect source record Inspected artifact
attributes.review.artifact_sha256 df36e4772562329550c924004a546812c8d76149a1b5238e4d0b0cfa8a82bb5c Source metadata	Frozen but Not Always Accessible: A Representation Analysis of Genomic Language Models Original source ↗ No field-specific location recorded Version: arXiv:2608.05329v1, submitted 2026-08-05 Retrieved: 2026-10-11T00:51:27Z	catalogued No individual claim review recorded Source has a recorded evidence concern. Consult its source page before using the claim. Audit details Field: `attributes.review.artifact_sha256` Source artifact SHA-256: `df36e4772562329550c924004a546812c8d76149a1b5238e4d0b0cfa8a82bb5c` Hash scope: Hash scope not separately documented; inspect source record Inspected artifact
Review: date 2026-10-11 Source metadata	Frozen but Not Always Accessible: A Representation Analysis of Genomic Language Models Original source ↗ No field-specific location recorded Version: arXiv:2608.05329v1, submitted 2026-08-05 Retrieved: 2026-10-11T00:51:27Z	catalogued No individual claim review recorded Source has a recorded evidence concern. Consult its source page before using the claim. Audit details Field: `attributes.review.date` Source artifact SHA-256: `df36e4772562329550c924004a546812c8d76149a1b5238e4d0b0cfa8a82bb5c` Hash scope: Hash scope not separately documented; inspect source record Inspected artifact
Review method 2 values source-hash-verification ai-assisted-source-review Source metadata	Frozen but Not Always Accessible: A Representation Analysis of Genomic Language Models Original source ↗ No field-specific location recorded Version: arXiv:2608.05329v1, submitted 2026-08-05 Retrieved: 2026-10-11T00:51:27Z	catalogued No individual claim review recorded Source has a recorded evidence concern. Consult its source page before using the claim. Audit details Field: `attributes.review.method` Source artifact SHA-256: `df36e4772562329550c924004a546812c8d76149a1b5238e4d0b0cfa8a82bb5c` Hash scope: Hash scope not separately documented; inspect source record Inspected artifact

How to read the evidence

Sources and history

Release 2026-10-22-1b7cf027e4f6 · Record review: source checked

arXiv preprint v1 probing frozen embeddings of five genomic language models against fine-tuned references on the Nucleotide Transformer benchmark, Genomic Benchmark and two Genomics Long-Range Benchmark variant-effect tasks.

Release history

No supporting source is linked yet.

Read original source

Download this release (gzip)

Technical metadata and extraction receipts

Stable ID: gfm2-20261011-datta2026-source

areas: dna-genomes
contexts: research
url: https://arxiv.org/abs/2608.05329v1
artifact url: https://arxiv.org/pdf/2608.05329v1
artifact sha256: df36e4772562329550c924004a546812c8d76149a1b5238e4d0b0cfa8a82bb5c
retrieved at: 2026-10-11T00:51:27Z
publication status: preprint
version: arXiv:2608.05329v1, submitted 2026-08-05
licence: arXiv.org perpetual non-exclusive license 1.0 (no redistribution of the bytes)
media type: application/pdf
source locator: Tables 1-9
evidence concerns: source id: gfm2-20261011-datta2026-source; message: Table 1 prints the same frozen value and spread (.5252+-.0262) for NT on both 'Enhancer' and 'Enhancer Types', while the fine-tuned references differ (.548 and .424). One of the two frozen cells is likely a transcription error in the source; both are kept as printed and should not support comparison until the authors clarify.; source locator: Table 1, column 'NT', rows 'Enhancer' and 'Enhancer Types'; artifact sha256: df36e4772562329550c924004a546812c8d76149a1b5238e4d0b0cfa8a82bb5c; review date: 2026-10-11; source id: gfm2-20261011-datta2026-source; message: Table 1 and Table 4 disagree for HyenaDNA Splice Acceptor: Table 1 prints .4752±.0108 (recovery 50), while Table 4 prints 0.7308 ± 0.0033 (validation-selected) and 0.7360 ± 0.0052 (diagnostic grid-best). Five other HyenaDNA cells of Table 1 equal Table 4's grid-best values, but Splice Donor (.7247±.0087) equals the validation-selected value instead. Table 3 and Table 4 also name different grid-best readouts for HyenaDNA Splice All (MLP+Mean against MLP+Attn). Table 1 values are kept as printed.; source locator: Table 1, column 'HyenaDNA', rows 'Splice Acceptor' and 'Splice Donor'; Table 4; Table 3, row 'Splice All'; artifact sha256: df36e4772562329550c924004a546812c8d76149a1b5238e4d0b0cfa8a82bb5c; review date: 2026-10-11
review: method: source-hash-verification; ai-assisted-source-review; reviewer: claude; reviewer note: Claude (Opus 5.5) review agent reviewer-gfm-a, independent of the extractor lit-gfm-2; no human review claimed; date: 2026-10-11; artifact sha256: df36e4772562329550c924004a546812c8d76149a1b5238e4d0b0cfa8a82bb5c; retrieval url: https://arxiv.org/pdf/2608.05329v1; note: Fresh download matches the pinned SHA-256. arXiv non-exclusive distribution licence; no copy archived.

Related records

Suggest a correction