Datasets:

JacobLinCool
/

AuraBench

@@ -783,344 +783,6 @@ tags:
 - signal-processing
 - exam
 ---
-<!-- a2b:header:start -->
-Machine-gradable exam benchmarks produced by **any-to-bench**. Each subset is one
-exam: the viewer table shows one row per answerable question (figures embedded);
-the raw, byte-faithful bundle lives under `<subset>/bundle/` — `exam.json`
-(structured paper), `answer_schema.json` (strict JSON Schema an answer sheet must
-satisfy), `grading.json` (deterministic rules + judge rubrics), `manifest.json`
-(provenance), `assets/` (figures), and, when present, `resources/` (the public
-solver corpus). The entire resource folder is public solver input; check it for
-secrets, answer material, and build artifacts before upload.
-## Usage
-Benchmark any model against an exam:
-```bash
-a2b download JacobLinCool/AuraBench --name <subset> -o bundle
-a2b solve bundle --model <your-model> -o answers.json
-a2b grade bundle answers.json -o report.json
-```
-Or load the question table directly:
-```python
-from datasets import load_dataset
-ds = load_dataset("JacobLinCool/AuraBench", "<subset>", split="test")
-```
-## ⚠️ Answer key included
-`<subset>/bundle/grading.json` contains the full answer key and scoring rubrics.
-If you benchmark models against this dataset, keep it out of training corpora.
-<!-- a2b:header:end -->
-<!-- a2b:bundle:waveform:start -->
-## waveform — AuraBench: Acoustic Information from Deterministic Visualizations
-| | |
-|---|---|
-| Subject | Image-only acoustic understanding |
-| Language | en |
-| Questions | 48 (48 auto-graded, 0 LLM-judged) |
-| Total points | 48 |
-| Question types | single_choice ×48 |
-| Created | 2026-08-24 |
-| Sources | `questions.yaml` `sha256:f0c77ec64833…`<br>`representations.yaml` `sha256:8a403b79e031…`<br>`source_checksums.json` `sha256:fdc96957d99e…`<br>`public_layout.yaml` `sha256:4cf6960f551c…` |
-`a2b download JacobLinCool/AuraBench --name waveform -o bundle`
-<!-- a2b:bundle:waveform:end -->
-<!-- a2b:bundle:stft:start -->
-## stft — AuraBench: Acoustic Information from Deterministic Visualizations
-| | |
-|---|---|
-| Subject | Image-only acoustic understanding |
-| Language | en |
-| Questions | 48 (48 auto-graded, 0 LLM-judged) |
-| Total points | 48 |
-| Question types | single_choice ×48 |
-| Created | 2026-08-24 |
-| Sources | `questions.yaml` `sha256:f0c77ec64833…`<br>`representations.yaml` `sha256:8a403b79e031…`<br>`source_checksums.json` `sha256:fdc96957d99e…`<br>`public_layout.yaml` `sha256:4cf6960f551c…` |
-`a2b download JacobLinCool/AuraBench --name stft -o bundle`
-<!-- a2b:bundle:stft:end -->
-<!-- a2b:bundle:mel:start -->
-## mel — AuraBench: Acoustic Information from Deterministic Visualizations
-| | |
-|---|---|
-| Subject | Image-only acoustic understanding |
-| Language | en |
-| Questions | 48 (48 auto-graded, 0 LLM-judged) |
-| Total points | 48 |
-| Question types | single_choice ×48 |
-| Created | 2026-08-24 |
-| Sources | `questions.yaml` `sha256:f0c77ec64833…`<br>`representations.yaml` `sha256:8a403b79e031…`<br>`source_checksums.json` `sha256:fdc96957d99e…`<br>`public_layout.yaml` `sha256:4cf6960f551c…` |
-`a2b download JacobLinCool/AuraBench --name mel -o bundle`
-<!-- a2b:bundle:mel:end -->
-<!-- a2b:bundle:cqt:start -->
-## cqt — AuraBench: Acoustic Information from Deterministic Visualizations
-| | |
-|---|---|
-| Subject | Image-only acoustic understanding |
-| Language | en |
-| Questions | 48 (48 auto-graded, 0 LLM-judged) |
-| Total points | 48 |
-| Question types | single_choice ×48 |
-| Created | 2026-08-24 |
-| Sources | `questions.yaml` `sha256:f0c77ec64833…`<br>`representations.yaml` `sha256:8a403b79e031…`<br>`source_checksums.json` `sha256:fdc96957d99e…`<br>`public_layout.yaml` `sha256:4cf6960f551c…` |
-`a2b download JacobLinCool/AuraBench --name cqt -o bundle`
-<!-- a2b:bundle:cqt:end -->
-<!-- a2b:bundle:multires_stft:start -->
-## multires_stft — AuraBench: Acoustic Information from Deterministic Visualizations
-| | |
-|---|---|
-| Subject | Image-only acoustic understanding |
-| Language | en |
-| Questions | 48 (48 auto-graded, 0 LLM-judged) |
-| Total points | 48 |
-| Question types | single_choice ×48 |
-| Created | 2026-08-24 |
-| Sources | `questions.yaml` `sha256:f0c77ec64833…`<br>`representations.yaml` `sha256:8a403b79e031…`<br>`source_checksums.json` `sha256:fdc96957d99e…`<br>`public_layout.yaml` `sha256:4cf6960f551c…` |
-`a2b download JacobLinCool/AuraBench --name multires_stft -o bundle`
-<!-- a2b:bundle:multires_stft:end -->
-<!-- a2b:bundle:reassigned:start -->
-## reassigned — AuraBench: Acoustic Information from Deterministic Visualizations
-| | |
-|---|---|
-| Subject | Image-only acoustic understanding |
-| Language | en |
-| Questions | 48 (48 auto-graded, 0 LLM-judged) |
-| Total points | 48 |
-| Question types | single_choice ×48 |
-| Created | 2026-08-24 |
-| Sources | `questions.yaml` `sha256:f0c77ec64833…`<br>`representations.yaml` `sha256:8a403b79e031…`<br>`source_checksums.json` `sha256:fdc96957d99e…`<br>`public_layout.yaml` `sha256:4cf6960f551c…` |
-`a2b download JacobLinCool/AuraBench --name reassigned -o bundle`
-<!-- a2b:bundle:reassigned:end -->
-<!-- a2b:bundle:scalogram:start -->
-## scalogram — AuraBench: Acoustic Information from Deterministic Visualizations
-| | |
-|---|---|
-| Subject | Image-only acoustic understanding |
-| Language | en |
-| Questions | 48 (48 auto-graded, 0 LLM-judged) |
-| Total points | 48 |
-| Question types | single_choice ×48 |
-| Created | 2026-08-24 |
-| Sources | `questions.yaml` `sha256:f0c77ec64833…`<br>`representations.yaml` `sha256:8a403b79e031…`<br>`source_checksums.json` `sha256:fdc96957d99e…`<br>`public_layout.yaml` `sha256:4cf6960f551c…` |
-`a2b download JacobLinCool/AuraBench --name scalogram -o bundle`
-<!-- a2b:bundle:scalogram:end -->
-<!-- a2b:bundle:erb:start -->
-## erb — AuraBench: Acoustic Information from Deterministic Visualizations
-| | |
-|---|---|
-| Subject | Image-only acoustic understanding |
-| Language | en |
-| Questions | 48 (48 auto-graded, 0 LLM-judged) |
-| Total points | 48 |
-| Question types | single_choice ×48 |
-| Created | 2026-08-24 |
-| Sources | `questions.yaml` `sha256:f0c77ec64833…`<br>`representations.yaml` `sha256:8a403b79e031…`<br>`source_checksums.json` `sha256:fdc96957d99e…`<br>`public_layout.yaml` `sha256:4cf6960f551c…` |
-`a2b download JacobLinCool/AuraBench --name erb -o bundle`
-<!-- a2b:bundle:erb:end -->
-<!-- a2b:bundle:periodicity:start -->
-## periodicity — AuraBench: Acoustic Information from Deterministic Visualizations
-| | |
-|---|---|
-| Subject | Image-only acoustic understanding |
-| Language | en |
-| Questions | 48 (48 auto-graded, 0 LLM-judged) |
-| Total points | 48 |
-| Question types | single_choice ×48 |
-| Created | 2026-08-24 |
-| Sources | `questions.yaml` `sha256:f0c77ec64833…`<br>`representations.yaml` `sha256:8a403b79e031…`<br>`source_checksums.json` `sha256:fdc96957d99e…`<br>`public_layout.yaml` `sha256:4cf6960f551c…` |
-`a2b download JacobLinCool/AuraBench --name periodicity -o bundle`
-<!-- a2b:bundle:periodicity:end -->
-<!-- a2b:bundle:modulation:start -->
-## modulation — AuraBench: Acoustic Information from Deterministic Visualizations
-| | |
-|---|---|
-| Subject | Image-only acoustic understanding |
-| Language | en |
-| Questions | 48 (48 auto-graded, 0 LLM-judged) |
-| Total points | 48 |
-| Question types | single_choice ×48 |
-| Created | 2026-08-24 |
-| Sources | `questions.yaml` `sha256:f0c77ec64833…`<br>`representations.yaml` `sha256:8a403b79e031…`<br>`source_checksums.json` `sha256:fdc96957d99e…`<br>`public_layout.yaml` `sha256:4cf6960f551c…` |
-`a2b download JacobLinCool/AuraBench --name modulation -o bundle`
-<!-- a2b:bundle:modulation:end -->
-<!-- a2b:bundle:source_filter:start -->
-## source_filter — AuraBench: Acoustic Information from Deterministic Visualizations
-| | |
-|---|---|
-| Subject | Image-only acoustic understanding |
-| Language | en |
-| Questions | 48 (48 auto-graded, 0 LLM-judged) |
-| Total points | 48 |
-| Question types | single_choice ×48 |
-| Created | 2026-08-24 |
-| Sources | `questions.yaml` `sha256:f0c77ec64833…`<br>`representations.yaml` `sha256:8a403b79e031…`<br>`source_checksums.json` `sha256:fdc96957d99e…`<br>`public_layout.yaml` `sha256:4cf6960f551c…` |
-`a2b download JacobLinCool/AuraBench --name source_filter -o bundle`
-<!-- a2b:bundle:source_filter:end -->
-<!-- a2b:bundle:prosody_voice:start -->
-## prosody_voice — AuraBench: Acoustic Information from Deterministic Visualizations
-| | |
-|---|---|
-| Subject | Image-only acoustic understanding |
-| Language | en |
-| Questions | 48 (48 auto-graded, 0 LLM-judged) |
-| Total points | 48 |
-| Question types | single_choice ×48 |
-| Created | 2026-08-24 |
-| Sources | `questions.yaml` `sha256:f0c77ec64833…`<br>`representations.yaml` `sha256:8a403b79e031…`<br>`source_checksums.json` `sha256:fdc96957d99e…`<br>`public_layout.yaml` `sha256:4cf6960f551c…` |
-`a2b download JacobLinCool/AuraBench --name prosody_voice -o bundle`
-<!-- a2b:bundle:prosody_voice:end -->
-<!-- a2b:bundle:waveform_mel:start -->
-## waveform_mel — AuraBench: Acoustic Information from Deterministic Visualizations
-| | |
-|---|---|
-| Subject | Image-only acoustic understanding |
-| Language | en |
-| Questions | 48 (48 auto-graded, 0 LLM-judged) |
-| Total points | 48 |
-| Question types | single_choice ×48 |
-| Created | 2026-08-24 |
-| Sources | `questions.yaml` `sha256:f0c77ec64833…`<br>`representations.yaml` `sha256:8a403b79e031…`<br>`source_checksums.json` `sha256:fdc96957d99e…`<br>`public_layout.yaml` `sha256:4cf6960f551c…` |
-`a2b download JacobLinCool/AuraBench --name waveform_mel -o bundle`
-<!-- a2b:bundle:waveform_mel:end -->
-<!-- a2b:bundle:mel_periodicity:start -->
-## mel_periodicity — AuraBench: Acoustic Information from Deterministic Visualizations
-| | |
-|---|---|
-| Subject | Image-only acoustic understanding |
-| Language | en |
-| Questions | 48 (48 auto-graded, 0 LLM-judged) |
-| Total points | 48 |
-| Question types | single_choice ×48 |
-| Created | 2026-08-24 |
-| Sources | `questions.yaml` `sha256:f0c77ec64833…`<br>`representations.yaml` `sha256:8a403b79e031…`<br>`source_checksums.json` `sha256:fdc96957d99e…`<br>`public_layout.yaml` `sha256:4cf6960f551c…` |
-`a2b download JacobLinCool/AuraBench --name mel_periodicity -o bundle`
-<!-- a2b:bundle:mel_periodicity:end -->
-<!-- a2b:bundle:speech_atlas:start -->
-## speech_atlas — AuraBench: Acoustic Information from Deterministic Visualizations
-| | |
-|---|---|
-| Subject | Image-only acoustic understanding |
-| Language | en |
-| Questions | 48 (48 auto-graded, 0 LLM-judged) |
-| Total points | 48 |
-| Question types | single_choice ×48 |
-| Created | 2026-08-24 |
-| Sources | `questions.yaml` `sha256:f0c77ec64833…`<br>`representations.yaml` `sha256:8a403b79e031…`<br>`source_checksums.json` `sha256:fdc96957d99e…`<br>`public_layout.yaml` `sha256:4cf6960f551c…` |
-`a2b download JacobLinCool/AuraBench --name speech_atlas -o bundle`
-<!-- a2b:bundle:speech_atlas:end -->
-<!-- a2b:bundle:acoustic_atlas:start -->
-## acoustic_atlas — AuraBench: Acoustic Information from Deterministic Visualizations
-| | |
-|---|---|
-| Subject | Image-only acoustic understanding |
-| Language | en |
-| Questions | 48 (48 auto-graded, 0 LLM-judged) |
-| Total points | 48 |
-| Question types | single_choice ×48 |
-| Created | 2026-08-24 |
-| Sources | `questions.yaml` `sha256:f0c77ec64833…`<br>`representations.yaml` `sha256:8a403b79e031…`<br>`source_checksums.json` `sha256:fdc96957d99e…`<br>`public_layout.yaml` `sha256:4cf6960f551c…` |
-`a2b download JacobLinCool/AuraBench --name acoustic_atlas -o bundle`
-<!-- a2b:bundle:acoustic_atlas:end -->
-<!-- a2b:bundle:control_no_image:start -->
-## control_no_image — AuraBench: Acoustic Information from Deterministic Visualizations
-| | |
-|---|---|
-| Subject | Image-only acoustic understanding |
-| Language | en |
-| Questions | 48 (48 auto-graded, 0 LLM-judged) |
-| Total points | 48 |
-| Question types | single_choice ×48 |
-| Created | 2026-08-24 |
-| Sources | `questions.yaml` `sha256:f0c77ec64833…`<br>`representations.yaml` `sha256:8a403b79e031…`<br>`source_checksums.json` `sha256:fdc96957d99e…`<br>`public_layout.yaml` `sha256:4cf6960f551c…` |
-`a2b download JacobLinCool/AuraBench --name control_no_image -o bundle`
-<!-- a2b:bundle:control_no_image:end -->
-<!-- a2b:bundle:control_blank_image:start -->
-## control_blank_image — AuraBench: Acoustic Information from Deterministic Visualizations
-| | |
-|---|---|
-| Subject | Image-only acoustic understanding |
-| Language | en |
-| Questions | 48 (48 auto-graded, 0 LLM-judged) |
-| Total points | 48 |
-| Question types | single_choice ×48 |
-| Created | 2026-08-24 |
-| Sources | `questions.yaml` `sha256:f0c77ec64833…`<br>`representations.yaml` `sha256:8a403b79e031…`<br>`source_checksums.json` `sha256:fdc96957d99e…`<br>`public_layout.yaml` `sha256:4cf6960f551c…` |
-`a2b download JacobLinCool/AuraBench --name control_blank_image -o bundle`
-<!-- a2b:bundle:control_blank_image:end -->
-<!-- a2b:bundle:control_mel_shuffled:start -->
-## control_mel_shuffled — AuraBench: Acoustic Information from Deterministic Visualizations
-| | |
-|---|---|
-| Subject | Image-only acoustic understanding |
-| Language | en |
-| Questions | 48 (48 auto-graded, 0 LLM-judged) |
-| Total points | 48 |
-| Question types | single_choice ×48 |
-| Created | 2026-08-25 |
-| Sources | `questions.yaml` `sha256:f0c77ec64833…`<br>`representations.yaml` `sha256:8a403b79e031…`<br>`source_checksums.json` `sha256:fdc96957d99e…`<br>`public_layout.yaml` `sha256:4cf6960f551c…` |
-`a2b download JacobLinCool/AuraBench --name control_mel_shuffled -o bundle`
-<!-- a2b:bundle:control_mel_shuffled:end -->
-<!-- aurabench:card:start -->
 # AuraBench Dataset Card
 ## Summary
@@ -1134,7 +796,8 @@ family.
 The shuffled-image control uses a deterministic semantic derangement within each
 compatible choice set: no image remains on its source question, no image crosses to a
 different option vocabulary, and every source image's ground-truth label differs from
-the target question's answer.
 Evaluated models receive image and text inputs only. The Any-to-Bench bundles do not
 contain audio files, transcripts, source filenames, corpus labels, or descriptive alt
@@ -1159,6 +822,26 @@ is shuffled; every question has its own pinned option permutation; and correct-o
 positions are counterbalanced for comparable choice sets. The same audited public
 layout is used across every image condition to preserve paired evaluation.
 ## Image conditions
 Atomic conditions are waveform, linear STFT, mel spectrogram, CQT,
@@ -1216,4 +899,3 @@ and any required image assets. `aurabench.json` records the image-to-question ma
 including the deliberate permutation used by the shuffled-image control, using only
 opaque public IDs. `config/public_layout.yaml` pins the reproducible question and option
 permutations and records the answer-shortcut audit score.
-<!-- aurabench:card:end -->

 - signal-processing
 - exam
 ---
 # AuraBench Dataset Card
 ## Summary
 The shuffled-image control uses a deterministic semantic derangement within each
 compatible choice set: no image remains on its source question, no image crosses to a
 different option vocabulary, and every source image's ground-truth label differs from
+the target question's answer. Its taker-visible exam identity matches the ordinary Mel
+condition, so the model is not told that image labels have been permuted.
 Evaluated models receive image and text inputs only. The Any-to-Bench bundles do not
 contain audio files, transcripts, source filenames, corpus labels, or descriptive alt
 positions are counterbalanced for comparable choice sets. The same audited public
 layout is used across every image condition to preserve paired evaluation.
+## Evaluation integrity
+The published Any-to-Bench bundles include `grading.json` so results can be reproduced.
+That file must never be staged into the taker's workspace. For agentic CLI takers, a private
+HOME, disabled web search, blocked shell network, and Codex `workspace-write` mode are still
+insufficient by themselves: `workspace-write` permits reads elsewhere on the host, including
+the source bundle and previous run outputs.
+The reference AuraBench runner therefore applies an outer OS-level filesystem policy. The
+taker can read and write its current ephemeral a2b run root and can read the pinned Codex
+runtime, but it cannot read the project directory, other a2b workspaces, mounted volumes, or
+network mounts, and cannot write outside its current run root. OS canaries verify these
+boundaries before benchmark execution. See `scripts/hermetic-bin/codex` and
+`scripts/run_matrix`.
+The corrected reference matrix contains 38 condition-effort pairs and 76 model runs, all with
+48/48 coverage and zero schema errors. The hidden shuffled-Mel control scores 7/48 and 10/48
+for sol low/medium, compared with the former impossible 48/48 result obtained when host answer
+keys were readable. Detailed results and SHA-256 digests are in `results/summary.json`.
 ## Image conditions
 Atomic conditions are waveform, linear STFT, mel spectrogram, CQT,
 including the deliberate permutation used by the shuffled-image control, using only
 opaque public IDs. `config/public_layout.yaml` pins the reproducible question and option
 permutations and records the answer-shortcut audit score.