BIRD (BIg Bench for LaRge-scale Database Grounded Text-to-SQL Evaluation) represents a pioneering, cross-domain dataset that examines the impact of extensive database contents on text-to-SQL parsing. BIRD contains over 12,751 unique question-SQL pairs and 95 big databases with a total size of 33.4 GB. It also covers more than 37 professional domains, such as blockchain, hockey, healthcare and education, etc.

nlile/BIRD-bench dataset hosted on Hugging Face and contributed by the HF Datasets community

madoss/bird-bench-questions dataset hosted on Hugging Face and contributed by the HF Datasets community

BIRD-SQL

Data from BIRD-SQL benchmark dev set (last release Jul 3, 2024). Ref: https://bird-bench.github.io

link to original dataset: https://bird-bench.github.io/ Li, J., Hui, B., Qu, G., Yang, J., Li, B., Li, B., Wang, B., Qin, B., Geng, R., Huo, N. and Zhou, X., 2024. Can llm already serve as a database interface? a big bench for large-scale database grounded text-to-sqls. Advances in Neural Information Processing Systems, 36.

BIRD-SQL Dataset

BIRD (BIg Bench for LaRge-scale Database Grounded Text-to-SQL Evaluation) is a comprehensive text-to-SQL dataset featuring realistic databases and complex queries across multiple domains. This dataset maintains the exact original BIRD format and field names.

  Dataset Statistics

Train: ~9,400 examples Validation: ~1,500 examples
Total: ~10,900 examples Databases: 80+ realistic databases

  Quick Start

from datasets import load_dataset

dataset =… See the full description on the dataset page: https://huggingface.co/datasets/Sudnya/bird-sql.

Update 2025-06-08

We release the full version of BIRD-Critic-PG, a dataset containing 530 high-quality user issues focused on real-world PostgreSQL database applications. The schema file is include in the code repository https://github.com/bird-bench/BIRD-CRITIC-1/blob/main/baseline/data/post_schema.jsonl

  BIRD-CRITIC-1.0-PG

BIRD-Critic is the first SQL debugging benchmark designed to answer a critical question: Can large language models (LLMs) fix user issues in… See the full description on the dataset page: https://huggingface.co/datasets/birdsql/bird-critic-1.0-postgresql.

Topic: In many aquatic environments, daily changes in oxygen concentrations occur frequently. When oxygen levels drop to hypoxic levels, many fishes respond with aquatic surface respiration, during which they obtain oxygen by skimming the upper, oxygenated surface layer. By increasing time at the surface, fish become more vulnerable to fish-eating birds. We explored these cascading effects in a sulfidic spring system that harbors the endemic sulfur molly (Poecilia sulphuraria) as prey species and several fish-eating bird species. Study design: All observations stem from the sulphuric spring complex ‘Baños del Azufre’ in southern Mexico. To explore whether the observed diurnal differences in fish behavior are driven by physicochemical water conditions and how they link to predator activity, we conducted two field surveys and one laboratory experiment. During one field season (subsequently termed field survey I), we first quantified how variable physicochemical water conditions were throughout the day. In another field season (subsequently termed field survey II), we investigated the link between fish’s behavior and predatory bird activity. To link observations from both surveys, they were carried out at the same location and followed the same regimen by sampling each morning (07:00 – 09:30), midday (12:00 – 14:30) and afternoon (16:00 – 18:30) for six subsequent days. Surveys were matched for season (i.e., end of dry season; I: 12–17 April 2019, II: 05–10 May 2016) and we subsequently verified that no major deviations from the temperature trend occurred between years (see DATA_FieldSurvey_SI_Temp). DATA_FieldSurvey_WaterChem: Physico-chemical water parametersWe took multiple measurements of water temperature, dissolved oxygen (DO), pH and electrical conductivity (EC), and – due to logistical constraints – one sulfide sample along a cross section of the sulfidic stream. To capture variation introduced through flow regime and/or stratification within the water column, measurements were taken according to a transect grid (i.e., one sample each meter starting at 0.5 m away from the stream bench and at depths of 0.05, 0.25, and 0.5 m as water levels would allow). With a width of 5.6 m and a depth of less than 0.8 m at the site, this resulted in 10–11 subsamples per sampling. DATA_FieldSurvey_Fish: Sulphur mollies’ diving behavior We assessed fish behavior using a focal animal sampling approach. Twenty diving fish of similar size were chosen randomly from a focal shoal and observed from the moment they initiated diving until they resurfaced to calculate the mean dive duration. DATA_FieldSurvey_Birds: Predatory bird activityWe recorded all sightings of piscivorous birds in the predefined study area within a 30-min period. For each bird, we determined species (to lowest feasible taxonomical level), entry and exit times, as well as the number of attacks launched. We calculated presence time as the mean time piscivorous birds spent in the study area (excluding mere fly-throughs, i.e., flying through the transect without landing or attacking) and attacks as the total number of bird attacks launched in a sampling period.DATA_FieldSurvey_SI_Temp: temperature monitoringAir and water temperatures were continuously monitored for the duration of the 2016 field survey. Two HOBO Pendant data loggers (Onset Computer Corp.) were placed in a shaded area at the upstream edge of the study area: one into the water at a depth of 15 centimeters, another 1.5 meters above the water. Temperatures were recorded at a ten-minute interval. For the purpose of this study, we only analyzed temperatures recorded during observational periods of birds and fish (i.e. ~1h per sampling period). DATA_LabExp_ASR: ASR tendency in response to hypoxiaWe exposed fish to dissolved oxygen concentrations ranging from near-anoxic to normoxic conditions (0.6 – 5.1 mg/L DO) in a laboratory setting. We tested a total of 27 groups (5 adult individuals each; n = 135). A trial lasted 10 minutes, but we only analyzed the last 5 minutes to ensue fish had recovered from handling and resumed swimming. Data shown here is group-pooled. We quantified the cumulative time spent at the surface by all five fish and calculated a percentage surface time. We also assessed the total number of dives and mean dive duration performed by each group.Lukas2021_WaterChem.R (see Lukas2021_WaterChem_Report for PDF version)All analyses were performed in R (R Core Team, 2020, version 4.0.2).

bibtex ref @article{li2024can, title={Can llm already serve as a database interface? a big bench for large-scale database grounded text-to-sqls}, author={Li, Jinyang and Hui, Binyuan and Qu, Ge and Yang, Jiaxi and Li, Binhua and Li, Bowen and Wang, Bailin and Qin, Bowen and Geng, Ruiying and Huo, Nan and others}, journal={Advances in Neural Information Processing Systems}, volume={36}, year={2024} }

The Aerospace Components Test Bench market is experiencing robust growth, projected to reach a market size of $381 million in 2025, exhibiting a Compound Annual Growth Rate (CAGR) of 6.6% from 2019 to 2033. This expansion is driven by several key factors. Firstly, the increasing demand for advanced aircraft and stricter safety regulations necessitate rigorous testing of components, fueling the demand for sophisticated test benches. Secondly, technological advancements leading to the development of more accurate and efficient test systems are contributing to market growth. Furthermore, the rising investments in research and development within the aerospace industry, coupled with the growing adoption of automated testing procedures, are further propelling market expansion. The diverse application segments, encompassing aircraft manufacturers, airlines, and maintenance companies, underscore the market's broad reach and potential for sustained growth. The major segments—Helicopter Gearbox Test Benches, Dynamometric Benches for Aircraft Tires, and Iron Bird systems—represent significant revenue contributors. Regional analysis reveals a geographically diverse market. North America and Europe currently hold significant market shares, driven by established aerospace industries and substantial investments in infrastructure. However, the Asia-Pacific region is anticipated to witness substantial growth in the coming years, fueled by increasing aircraft production and a developing aerospace sector. Competitive landscape analysis shows a mix of established players and emerging companies, leading to increased innovation and a wider range of solutions. The presence of both large multinational corporations and specialized niche players highlights the diverse nature of the market and the opportunities for both established and new entrants. Ongoing technological advancements, such as the integration of artificial intelligence and machine learning, are likely to further reshape the market dynamics in the coming years.

Text2SQL Workflow Trace

  Dataset Description

This dataset contains workflow traces for Text-to-SQL tasks, capturing the intermediate steps of translating natural language queries to executable SQL. It was used as input trace for the research presented in the paper:"HEXGEN-TEXT2SQL: Optimizing LLM Inference Request Scheduling for Agentic Text-to-SQL Workflow" (arXiv:2505.05286). The end-to-end Text-to-SQL queries collected in the dataset are from BIRD bench, and the trace… See the full description on the dataset page: https://huggingface.co/datasets/fredpeng/Text2SQL_Workflow_Trace.

Original provider: Wake Forest University

Dataset credits: National Science Foundation

Abstract: Satellite telemetry was used to identify the foraging distributions of two congeneric species of albatrosses that nest in the tropics/subtropics. Breeding black-footed albatross (Phoebastria nigripes) and Laysan albatross (Phoebastria immutabilis) nesting in Tern Island (Northwest Hawaiian Islands) and tracked during the 1998 breeding season (January - June) performed foraging trips to continental shelves off North America. Black-footed albatross made long trips to the west coast of North America (British Columbia to California). Laysan albatross traveled primarily to the north of the Hawaiian Islands, and reached the waters of the Aleutian Islands and the Gulf of Alaska. These albatross species mixed short and long trips during the chick-rearing period (February - June), but engaged in short foraging trips during the brooding period (within 18 days after chick hatched, January - February).

In 1999, the breeding success of both albatross species was depressed, with a large-scale failure for the Laysan albatross. Out of nine black-footed albatross tracked, two chicks died during this study. Out of sixteen Laysan albatross tracked, the eggs of seven birds did not hatch and eight chicks died during the tracking study. Due to this massive breeding failure, the satellite tracked birds abandoned their colony and dispersed widely across the North Pacific Ocean. Thus, the 1998 (central-place foraging) and 1999 (dispersal from colonies) tracking data should be considered separately.

Purpose: The 1999 data provided information on albatross movements during a year of depressed reproductive success, when many birds abandoned the colony. An understanding on the interplay between the distribution and the reproductive success of North Pacific albatrosses has important implications for assessing how oceanographic variability influences their population dynamics.

We thank C. Alexander, L. Carsten, P. Fernández, F. Juola, P. Sievert, A. Viggiano and S. Wang for assistance in the field, and the U.S. Fish and Wildlife Service for logistical support. This research was funded by National Science Foundation grant DEB 9629539 to D. Anderson.

Supplemental information: These albatross were tracked using PTT-100 Argos transmitters (Microwave Telemetry, Columbia, MD) operating at a 90-second repetition rate and programmed to operate on a 8:24 h ON:OFF duty cycle. Transmitter bench-tests before deployment revealed that the Argos location quality classes (lcs) had the following median position errors, expressed in kilometers: lc B (8.46), lc A (3.29), lc 0 (4.80), lc 1 (1.96), lc 2 (0.49), and lc 3 (0.26).

The low-quality class B locations were discarded because they mis-represented the telemetry tracks. Thus, this dataset includes 4635 high-quality locations (lc classes A or better) with median positional errors <4 km.