# ADS

Source: [ADS](<https://www.sota2.com/research/dataset/ads>)

Collection index: [llms.txt](<https://www.sota2.com/research/llms.txt>)

A research result applies to its stated dataset, split, metric and evaluation configuration. Compare like-for-like settings, preserve the metric's unit and direction, and distinguish paper-reported results from independent reproduction. The best result recorded here is not a claim of exhaustive, current global coverage.

## Benchmarks

Page 1 of 5. This is an index page, not the full collection.

- [Next index page](<https://www.sota2.com/research/dataset/ads.md?page=2>): Follow this link to discover more pages.

- [Temporal Action Segmentation on 50Salads](<https://www.sota2.com/research/sota/temporal-action-segmentation-on-50salads.md>): Temporal Action Segmentation; 50Salads.

- [Action Segmentation on 50Salads](<https://www.sota2.com/research/sota/action-segmentation-on-50salads.md>): Action Segmentation; 50Salads.

- [Indirect Prompt Injection Defense on ICLR Academic Reviews 84 defensive payloads / 252 prompts 2026](<https://www.sota2.com/research/sota/indirect-prompt-injection-defense-on-iclr-academic-reviews-84-defensive-payloads-252-prompts-2026.md>): Indirect Prompt Injection Defense; ICLR Academic Reviews 84 defensive payloads / 252 prompts 2026.

- [3D Object Detection on TJ4DRadSet (test)](<https://www.sota2.com/research/sota/3d-object-detection-on-tj4dradset-test.md>): 3D Object Detection; TJ4DRadSet (test).

- [Attention Operator Throughput on Llama 405B (128 Q-heads/8 KV-heads/128 Head-dimension) 3.1](<https://www.sota2.com/research/sota/attention-operator-throughput-on-llama-405b-128-q-heads-8-kv-heads-128-head-dimension-3-1.md>): Attention Operator Throughput; Llama 405B (128 Q-heads/8 KV-heads/128 Head-dimension) 3.1.

- [Spreadsheet Manipulation and Reasoning on SpreadsheetBench held-out (test)](<https://www.sota2.com/research/sota/spreadsheet-manipulation-and-reasoning-on-spreadsheetbench-held-out-test.md>): Spreadsheet Manipulation and Reasoning; SpreadsheetBench held-out (test).

- [Long-term Action Anticipation on 50 Salads](<https://www.sota2.com/research/sota/long-term-action-anticipation-on-50-salads.md>): Long-term action anticipation; 50 Salads.

- [Infrared-Visible Image Fusion on RoadScene (test)](<https://www.sota2.com/research/sota/infrared-visible-image-fusion-on-roadscene-test.md>): Infrared-Visible Image Fusion; RoadScene (test).

- [Determiner Productivity Evaluation on Manchester child dyads (test)](<https://www.sota2.com/research/sota/determiner-productivity-evaluation-on-manchester-child-dyads-test.md>): Determiner Productivity Evaluation; Manchester child dyads (test).

- [Classification on Roadside LiDAR dataset (test)](<https://www.sota2.com/research/sota/classification-on-roadside-lidar-dataset-test.md>): Classification; Roadside LiDAR dataset (test).

- [Cross-Position Human Activity Recognition on DSADS](<https://www.sota2.com/research/sota/cross-position-human-activity-recognition-on-dsads.md>): Cross-Position Human Activity Recognition; DSADS.

- [Infrared and visible image fusion on RoadScene](<https://www.sota2.com/research/sota/infrared-and-visible-image-fusion-on-roadscene.md>): Infrared and visible image fusion; RoadScene.

- [Attention Operator Throughput on Llama2 7B (32 Q-heads/32 KV-heads/128 Head-dimension)](<https://www.sota2.com/research/sota/attention-operator-throughput-on-llama2-7b-32-q-heads-32-kv-heads-128-head-dimension.md>): Attention Operator Throughput; Llama2 7B (32 Q-heads/32 KV-heads/128 Head-dimension).

- [CTR Prediction on TaobaoAds](<https://www.sota2.com/research/sota/ctr-prediction-on-taobaoads.md>): CTR Prediction; TaobaoAds.

- [Passage-pair linking on Literary analysis corpus (Goodreads, Sparknotes, Litcharts)](<https://www.sota2.com/research/sota/passage-pair-linking-on-literary-analysis-corpus-goodreads-sparknotes-litcharts.md>): Passage-pair linking; Literary analysis corpus (Goodreads, Sparknotes, Litcharts).

- [Road Scene on RoadSceneBench](<https://www.sota2.com/research/sota/road-scene-on-roadscenebench.md>): Road Scene; RoadSceneBench.

- [3D Hand Pose Estimation on AssemblyHands D\_syn -\> D\_ah (Headset2)](<https://www.sota2.com/research/sota/3d-hand-pose-estimation-on-assemblyhands-d_syn-d_ah-headset2.md>): 3D Hand Pose Estimation; AssemblyHands D\_syn -\> D\_ah (Headset2).

- [3D Hand Pose Estimation on AssemblyHands Headset1 (D\_syn -\> D\_ah)](<https://www.sota2.com/research/sota/3d-hand-pose-estimation-on-assemblyhands-headset1-d_syn-d_ah.md>): 3D Hand Pose Estimation; AssemblyHands Headset1 (D\_syn -\> D\_ah).

- [Visual Localization on 7Scenes Heads](<https://www.sota2.com/research/sota/visual-localization-on-7scenes-heads.md>): Visual Localization; 7Scenes Heads.

- [Recommendation on Goodreads (test)](<https://www.sota2.com/research/sota/recommendation-on-goodreads-test.md>): Recommendation; Goodreads (test).

- [Link Prediction on Goodreads-LP](<https://www.sota2.com/research/sota/link-prediction-on-goodreads-lp.md>): Link Prediction; Goodreads-LP.

- [BEV Object Detection on TJ4DRadSet (test)](<https://www.sota2.com/research/sota/bev-object-detection-on-tj4dradset-test.md>): BEV Object Detection; TJ4DRadSet (test).

- [End-to-end throughput on Cloud Inference Workloads](<https://www.sota2.com/research/sota/end-to-end-throughput-on-cloud-inference-workloads.md>): End-to-end throughput; Cloud Inference Workloads.

- [Unlearning on Goodreads-NC](<https://www.sota2.com/research/sota/unlearning-on-goodreads-nc.md>): Unlearning; Goodreads-NC.

- [Outlier Detection on InternetAds ADBench](<https://www.sota2.com/research/sota/outlier-detection-on-internetads-adbench.md>): Outlier Detection; InternetAds ADBench.

- [Relation-Conditioned Retrieval on Goodreads](<https://www.sota2.com/research/sota/relation-conditioned-retrieval-on-goodreads.md>): Relation-Conditioned Retrieval; Goodreads.

- [Spreadsheet Agent Skills on SpreadsheetBench (in-distribution)](<https://www.sota2.com/research/sota/spreadsheet-agent-skills-on-spreadsheetbench-in-distribution.md>): Spreadsheet Agent Skills; SpreadsheetBench (in-distribution).

- [Attention Operator Throughput on Qwen2.5 72B (64 Q-heads/8 KV-heads/128 Head-dimension)](<https://www.sota2.com/research/sota/attention-operator-throughput-on-qwen2-5-72b-64-q-heads-8-kv-heads-128-head-dimension.md>): Attention Operator Throughput; Qwen2.5 72B (64 Q-heads/8 KV-heads/128 Head-dimension).

- [Human Activity Recognition on DSADS Cross-person (Micro F1)](<https://www.sota2.com/research/sota/human-activity-recognition-on-dsads-cross-person-micro-f1.md>): Human Activity Recognition; DSADS Cross-person.

- [Peak Estimation on Goodreads per-genre](<https://www.sota2.com/research/sota/peak-estimation-on-goodreads-per-genre.md>): Peak Estimation; Goodreads per-genre.

- [Node Classification on Goodreads-NC (test)](<https://www.sota2.com/research/sota/node-classification-on-goodreads-nc-test.md>): Node Classification; Goodreads-NC (test).

- [Membership Inference Attack on SQuADShifts (test)](<https://www.sota2.com/research/sota/membership-inference-attack-on-squadshifts-test.md>): Membership Inference Attack; SQuADShifts (test).

- [Node Classification on Goodreads](<https://www.sota2.com/research/sota/node-classification-on-goodreads.md>): Node Classification; Goodreads.

- [Camera Pose Regression on 7Scenes Heads](<https://www.sota2.com/research/sota/camera-pose-regression-on-7scenes-heads.md>): Camera Pose Regression; 7Scenes Heads.

- [Spreadsheet Reasoning on SpreadsheetBench (test)](<https://www.sota2.com/research/sota/spreadsheet-reasoning-on-spreadsheetbench-test.md>): Spreadsheet Reasoning; SpreadsheetBench (test).

- [Overall Spreadsheet Performance on SpreadsheetBench and WikiTableQuestions Combined](<https://www.sota2.com/research/sota/overall-spreadsheet-performance-on-spreadsheetbench-and-wikitablequestions-combined.md>): Overall Spreadsheet Performance; SpreadsheetBench and WikiTableQuestions Combined.

- [Spreadsheet Task Execution on SpreadsheetBench Full](<https://www.sota2.com/research/sota/spreadsheet-task-execution-on-spreadsheetbench-full.md>): Spreadsheet Task Execution; SpreadsheetBench Full.

- [Spreadsheet Task Execution on SpreadsheetBench Verified (Held-out)](<https://www.sota2.com/research/sota/spreadsheet-task-execution-on-spreadsheetbench-verified-held-out.md>): Spreadsheet Task Execution; SpreadsheetBench Verified (Held-out).

- [Matrix Multiplication Performance on RTX 4090 (Sizes 2048-8192)](<https://www.sota2.com/research/sota/matrix-multiplication-performance-on-rtx-4090-sizes-2048-8192.md>): Matrix Multiplication Performance; RTX 4090 Matrix Workloads Matrix sizes: 2048 to 8192.

- [Spreadsheet Understanding and Manipulation on SpreadsheetBench Soft Restriction](<https://www.sota2.com/research/sota/spreadsheet-understanding-and-manipulation-on-spreadsheetbench-soft-restriction.md>): Spreadsheet Understanding and Manipulation; SpreadsheetBench Soft Restriction.

- [Action Segmentation on 50 Salads (eval)](<https://www.sota2.com/research/sota/action-segmentation-on-50-salads-eval.md>): Action Segmentation; 50 Salads (eval).

- [Temporal Action Segmentation on 50 Salads granularity (Eval)](<https://www.sota2.com/research/sota/temporal-action-segmentation-on-50-salads-granularity-eval.md>): Temporal Action Segmentation; 50 Salads granularity (Eval).

- [Temporal Action Segmentation on 50 Salads 65](<https://www.sota2.com/research/sota/temporal-action-segmentation-on-50-salads-65.md>): Temporal Action Segmentation; 50 Salads 65.

- [Dense anticipation mean over classes on 50salads (test)](<https://www.sota2.com/research/sota/dense-anticipation-mean-over-classes-on-50salads-test.md>): Dense anticipation mean over classes; 50salads (test).

- [Spreadsheet Understanding and Manipulation on SpreadsheetBench Hard Restriction](<https://www.sota2.com/research/sota/spreadsheet-understanding-and-manipulation-on-spreadsheetbench-hard-restriction.md>): Spreadsheet Understanding and Manipulation; SpreadsheetBench Hard Restriction.

- [Question Answering on SQuADShifts 1.0 (test)](<https://www.sota2.com/research/sota/question-answering-on-squadshifts-1-0-test.md>): Question Answering; SQuADShifts 1.0 (test).

- [Spreadsheet Tasks on SSB (SpreadsheetBench)](<https://www.sota2.com/research/sota/spreadsheet-tasks-on-ssb-spreadsheetbench.md>): Spreadsheet Tasks; SSB (SpreadsheetBench).

- [Dynamic 3D Reconstruction on vrheadset](<https://www.sota2.com/research/sota/dynamic-3d-reconstruction-on-vrheadset.md>): Dynamic 3D Reconstruction; vrheadset.

- [Spreadsheet Understanding on SpreadsheetLLM benchmark (298 queries)](<https://www.sota2.com/research/sota/spreadsheet-understanding-on-spreadsheetllm-benchmark-298-queries.md>): Spreadsheet Understanding; SpreadsheetLLM benchmark (298 queries).

- [Recommendation on Goodreads](<https://www.sota2.com/research/sota/recommendation-on-goodreads.md>): Recommendation; Goodreads.

- [Sub-workflow Execution on OS-Marathon Expense Report - Spreadsheet Environment](<https://www.sota2.com/research/sota/sub-workflow-execution-on-os-marathon-expense-report-spreadsheet-environment.md>): Sub-workflow Execution; OS-Marathon Expense Report - Spreadsheet Environment.

- [E-commerce Short Video Understanding on E-VAds 1.0 (test)](<https://www.sota2.com/research/sota/e-commerce-short-video-understanding-on-e-vads-1-0-test.md>): E-commerce Short Video Understanding; E-VAds 1.0 (test).

- [Activity Recognition on DSADS](<https://www.sota2.com/research/sota/activity-recognition-on-dsads.md>): Activity Recognition; DSADS.

- [Semantic Segmentation on RoadSeg Mnih (test)](<https://www.sota2.com/research/sota/semantic-segmentation-on-roadseg-mnih-test.md>): Semantic Segmentation; RoadSeg Mnih (test).

- [Mathematical Reasoning on Olympiads](<https://www.sota2.com/research/sota/mathematical-reasoning-on-olympiads.md>): Mathematical Reasoning; Olympiads.

- [Personalized Speech Enhancement on DNS Track 1: Headset 5 (test)](<https://www.sota2.com/research/sota/personalized-speech-enhancement-on-dns-track-1-headset-5-test.md>): Personalized Speech Enhancement; DNS Track 1: Headset 5 (test).

- [Action Segmentation on 50Salads (mid granularity)](<https://www.sota2.com/research/sota/action-segmentation-on-50salads-mid-granularity.md>): Action Segmentation; 50Salads mid granularity.

- [Spreadsheet Reasoning on SpreadsheetBench Verified](<https://www.sota2.com/research/sota/spreadsheet-reasoning-on-spreadsheetbench-verified.md>): Spreadsheet Reasoning; SpreadsheetBench Verified.

- [Program Repair on SpreadsheetBench (test)](<https://www.sota2.com/research/sota/program-repair-on-spreadsheetbench-test.md>): Program Repair; SpreadsheetBench (test).

- [Spreadsheet Understanding on SpreadsheetLLM benchmark](<https://www.sota2.com/research/sota/spreadsheet-understanding-on-spreadsheetllm-benchmark.md>): Spreadsheet Understanding; SpreadsheetLLM benchmark.

- [Road Topology on RoadSceneBench](<https://www.sota2.com/research/sota/road-topology-on-roadscenebench.md>): Road Topology; RoadSceneBench.

- [Traffic Condition on RoadSceneBench](<https://www.sota2.com/research/sota/traffic-condition-on-roadscenebench.md>): Traffic Condition; RoadSceneBench.

- [Lane Change Feasibility on RoadSceneBench](<https://www.sota2.com/research/sota/lane-change-feasibility-on-roadscenebench.md>): Lane Change Feasibility; RoadSceneBench.

- [Ego-lane Index on RoadSceneBench](<https://www.sota2.com/research/sota/ego-lane-index-on-roadscenebench.md>): Ego-lane Index; RoadSceneBench.

- [Lane Count on RoadSceneBench](<https://www.sota2.com/research/sota/lane-count-on-roadscenebench.md>): Lane Count; RoadSceneBench.

- [Sequential Recommendation on GoodReads Comics](<https://www.sota2.com/research/sota/sequential-recommendation-on-goodreads-comics.md>): Sequential Recommendation; GoodReads Comics.

- [Sequential Recommendation on GoodReads Children](<https://www.sota2.com/research/sota/sequential-recommendation-on-goodreads-children.md>): Sequential Recommendation; GoodReads Children.

- [Model Efficiency Evaluation on TaobaoAds for CTR Prediction](<https://www.sota2.com/research/sota/model-efficiency-evaluation-on-taobaoads-for-ctr-prediction.md>): Click-Through Rate Prediction; TaobaoAds.

- [Action Segmentation on 50 Salads Mid](<https://www.sota2.com/research/sota/action-segmentation-on-50-salads-mid.md>): Action Segmentation; 50 Salads Mid.

- [Long-horizon behavior prediction on OminiBehavior Ads](<https://www.sota2.com/research/sota/long-horizon-behavior-prediction-on-ominibehavior-ads.md>): Long-horizon behavior prediction; OminiBehavior Ads.

- [Sentiment Analysis on Goodreads](<https://www.sota2.com/research/sota/sentiment-analysis-on-goodreads.md>): Sentiment Analysis; Goodreads.

- [Book Review Rating on Goodreads](<https://www.sota2.com/research/sota/book-review-rating-on-goodreads.md>): Book Review Rating; Goodreads.

- [Trade Ordering for Price Stability on Simulated Trade Workloads 100 trials](<https://www.sota2.com/research/sota/trade-ordering-for-price-stability-on-simulated-trade-workloads-100-trials.md>): Trade Ordering for Price Stability; Simulated Trade Workloads 100 trials.

- [Spreadsheet Manipulation on SpreadsheetBench](<https://www.sota2.com/research/sota/spreadsheet-manipulation-on-spreadsheetbench.md>): Spreadsheet Manipulation; SpreadsheetBench.

- [Human Activity Recognition on DSADS](<https://www.sota2.com/research/sota/human-activity-recognition-on-dsads.md>): Human Activity Recognition; DSADS.

- [Time Series Anomaly Detection on DSADS](<https://www.sota2.com/research/sota/time-series-anomaly-detection-on-dsads.md>): Time Series Anomaly Detection; DSADS.

- [Action Segmentation on 50Salads (test)](<https://www.sota2.com/research/sota/action-segmentation-on-50salads-test.md>): Action Segmentation; 50Salads (test).

- [Cross-Scale and Cross-Domain Generalization on SpreadsheetBench and WikiTQ](<https://www.sota2.com/research/sota/cross-scale-and-cross-domain-generalization-on-spreadsheetbench-and-wikitq.md>): Cross-Scale and Cross-Domain Generalization; SpreadsheetBench and WikiTQ.

- [Sequential Recommendation on GoodReads](<https://www.sota2.com/research/sota/sequential-recommendation-on-goodreads.md>): Sequential Recommendation; GoodReads.

- [Traffic Law Violation Coverage on Chinese traffic law specification 82 violation goals Apollo ADS Derived from LawBreaker](<https://www.sota2.com/research/sota/traffic-law-violation-coverage-on-chinese-traffic-law-specification-82-violation-goals-apollo-ads-derived-from-lawbreaker.md>): Traffic Law Violation Coverage; Chinese traffic law specification 82 violation goals Apollo ADS Derived from LawBreaker.