Dataset Viewer
Auto-converted to Parquet Duplicate
input
stringlengths
159
1.58k
output
stringlengths
98
13.4k
timeseries
listlengths
1
23
eval_type
stringclasses
51 values
eval_metadata
stringlengths
48
534
You are a time series analysis expert. This is a metric called User Activity collected from Social Media with length of 81: <ts>. If a 'significant upward shift' is when the difference between second-half and first-half means is more than 15% of the total range, does User Activity qualify?
<think> metric: User Activity stats: max=79.95, min=13.43, first_half_mean=48.10, second_half_mean=28.76 === range = 79.95 - 13.43 = 66.52 shift = 28.76 - 48.10 = -19.34 |shift| = 19.34 threshold = 15% × 66.52 = 9.98 19.34 > 9.98 -> met answer: yes </think> Yes. User Activity's mean shifts significantly between halves...
[ [ 50.319246, 50.313482, 50.29619, 79.94737, 50.227021, 50.175144, 50.11174, 50.036807, 49.950346, 49.852357, 49.74284, 49.621795, 49.489221, 49.34512, 49.18949, 49.022332, 48.843646, 48.653432, 48.45169, 48.23842, 48.013621, 47....
segment_judgment
{"judgment_type": "half_mean_shift", "verdict": "yes", "length": 81, "tsevol": false}
In an Advertising system, there are 10 metrics: Engagement Rates on Ad Content is of length 256: <ts>; Campaign Reach is of length 256: <ts>; Attribution Metrics is of length 256: <ts>; Video Completion Rates is of length 256: <ts>; Social Media Ad Performance is of length 256: <ts>; Ad Frequency is of length 256...
<think> metric: Engagement Rates on Ad Content trend_segments=[(increase, 0, 255)] --- metric: Video Completion Rates trend_segments=[(decrease, 0, 255)] === threshold=2. comparison with Engagement Rates on Ad Content vs Video Completion Rates: segments count: 1 vs 1. seg 0: type increase ≠ decrease. FAIL. answer: diff...
[ [ 7.553012, 7.569893, 7.597657, 7.568316, 7.611129, 7.77914, 7.931604, 7.859522, 7.691188, 7.692333, 7.69418, 7.709237, 7.754803, 7.726179, 7.764055, 7.757782, 7.797529, 7.786007, 7.833648, 7.84219, 7.856823, 7.867752, 7.863...
correlation
{"metric_a": "Engagement Rates on Ad Content", "metric_b": "Video Completion Rates", "mode": "shape", "length": 256, "verdict": "different", "tsevol": false}
In an Internet of Things (IoT) system, there are 9 metrics: Smart Lighting Usage is of length 256: <ts>; Smart Home Metrics is of length 256: <ts>; Wearable Device Data is of length 256: <ts>; Health Monitoring Devices is of length 256: <ts>; Device Uptime/Downtime is of length 256: <ts>; Water Quality Sensors is...
<think> metric: Smart Lighting Usage start=-0.51, end=-0.25, amp=0.26, overall trend=increase trend_segments=[(increase, 0, 155), (decrease, 155, 255)] </think> From the perspective of the slope, the overall trend is increasing. The value of time series starts from around -0.51 and ends at around -0.25, with an overal...
[ [ -554.552834, -535.881737, -538.6504, -523.150733, -514.374829, -502.32157, -490.046021, -476.841906, -444.559462, -447.141986, -443.55328, -409.256695, -401.053788, -392.482284, -373.96781, -354.624706, -354.828838, -337.143491, -330.4947...
description
{"perspectives": ["trend"], "metric_index": 0, "length": 256, "tsevol": false}
In a Manufacturing system, there are 13 metrics: Maintenance Schedules is of length 256: <ts>; Downtime Logs is of length 256: <ts>; Equipment Utilization is of length 256: <ts>; Safety Incident Rates is of length 256: <ts>; Inventory Levels is of length 256: <ts>; Machine Performance is of length 256: <ts>; Qua...
<think> Let me look at the timeseries carefully. Looking at the Downtime Logs timeseries (256 data points). The minimum value of -1663187.95 occurs at index 159. answer: 159 </think> For Downtime Logs, the lowest value is at position 159.
[ [ -0.354261, -0.378228, -0.386434, -0.406975, -0.378872, -0.403445, -0.329792, -0.36726, -0.29401, -0.360344, -0.396578, -0.38435, -0.386309, -0.39152, -0.416849, -0.357787, -0.359897, -0.369881, -0.32867, -0.363386, -0.378996, ...
atomic_min_position
{"length": 256, "verdict": "159", "direction": "min", "tsevol": false}
You are a time series analysis expert. This is a metric called Undo Tablespace Usage collected from Oracle Database with length of 256: <ts>. Check whether Undo Tablespace Usage has a noticeable event around point 103 within the tolerance of 20 timesteps.
<think> metric: Undo Tablespace Usage local=[sudden decrease, start@(9, 389.57); sudden decrease, start@(76, 347.07); downward spike, start@(214, 288.97)] === threshold=20. anchor=103. Undo Tablespace Usage: |9-103|=94 > 20; |76-103|=27 > 20; |214-103|=111 > 20. FAIL. answer: no </think> I did not find any local event...
[ [ 393.637441, 393.185033, 392.732643, 392.280287, 391.827982, 391.375745, 390.923593, 390.471544, 390.019613, 389.567819, 386.523321, 383.478994, 380.434853, 377.390917, 374.347202, 371.303725, 368.260504, 367.627328, 368.492562, 370.19...
yes_no
{"target_point": 103, "metric": "Undo Tablespace Usage", "threshold": 20, "variant": "generic", "verdict": "no", "length": 256, "tsevol": false}
In a Finance system, there are 7 metrics: Volume Traded is of length 256: <ts>; Stock Prices is of length 256: <ts>; Dividend Yields is of length 256: <ts>; Exchange Rates is of length 256: <ts>; Market Indices is of length 256: <ts>; Cryptocurrency Prices is of length 256: <ts>; Inflation Rates is of length 256...
<think> metric: Cryptocurrency Prices trend_segments=[(increase, 0, 66), (decrease, 66, 255)] === durations: (increase, 0, 66): duration = 66 - 0 = 66 (decrease, 66, 255): duration = 255 - 66 = 189 aggregate: increase: 66 decrease: 189 decrease: 189 > 66 -> winner = decrease answer: decrease </think> Cryptocur...
[ [ 139.856541, 120.121509, 107.644389, 88.236778, 63.828286, 24.718832, 24.373145, -12.485533, -15.280939, -26.304056, -38.468855, -74.710935, -94.750586, -118.165415, -128.040678, -129.070112, -137.630422, -180.311948, -208.382899, -236...
duration_proportion
{"sub_type": "duration_comparison", "verdict": "decrease", "length": 256, "tsevol": false}
You are a time series analysis expert. This is a metric called Engagement Rates on Ad Content collected from Advertising with length of 256: <ts>. Where does the last local event appear in Engagement Rates on Ad Content?
<think> metric: Engagement Rates on Ad Content local=[sudden increase, pos=30; sudden decrease, pos=207] === sorted positions: 30, 207 latest = 207 answer: 207 </think> The last local event in Engagement Rates on Ad Content occurs at position 207.
[ [ 14.548736, 20.198397, 25.616435, 29.271793, 30.708193, 30.632328, 30.592903, 32.360959, 37.219863, 45.4038, 55.876729, 66.53443, 74.775586, 78.267556, 75.667336, 67.06917, 54.034208, 39.187897, 25.506018, 15.507503, 10.600412, ...
temporal_position
{"sub_type": "last_event", "verdict": "207", "length": 256, "tsevol": false}
In an Agriculture system, there are 5 metrics: Organic vs. Conventional Farming Metrics is of length 256: <ts>; Greenhouse Gas Emissions from Farming is of length 256: <ts>; Livestock Counts is of length 256: <ts>; Water Usage Efficiency is of length 256: <ts>; Feed Conversion Ratios is of length 256: <ts>. In Org...
<think> metric: Organic vs. Conventional Farming Metrics trend_segments=[(decrease, 0, 76), (increase, 76, 185), (decrease, 185, 255)] === intersecting [91, 242] with Organic vs. Conventional Farming Metrics trend segments: - (increase, 76, 185) ∩ [91, 242] = [91, 185] -> duration = 185 - 91 = 94 - (decrease, 185, ...
[ [ 94.265225, 90.324971, 92.742834, 75.458174, 81.782144, 69.383363, 72.336015, 66.628428, 65.453624, 55.699366, 53.079427, 47.571201, 51.64542, 41.359895, 38.733815, 31.472398, 26.50093, 22.971703, 19.393955, 19.435467, 17.776222, ...
segment_trend_dominance
{"verdict": "increase", "length": 256, "tsevol": false}
In a Manufacturing system, there are 13 metrics: Machine Performance is of length 256: <ts>; Production Line Efficiency is of length 256: <ts>; Lead Times is of length 256: <ts>; Inventory Turnover is of length 256: <ts>; Manufacturing Costs is of length 256: <ts>; Production Scheduling Metrics is of length 256: ...
<think> metric: Energy Usage in Production start=0.41, end=0.69, amp=0.28, overall trend=increase trend_segments=[(increase, 0, 67), (decrease, 67, 194), (increase, 194, 255)] </think> From the perspective of the slope, the overall trend is increasing. The value of time series starts from around 0.41 and ends at aroun...
[ [ 333.749906, 341.216045, 347.346193, 355.407747, 360.817207, 364.108558, 374.462195, 380.379256, 387.165879, 395.115385, 395.413345, 404.774045, 406.947132, 417.065661, 420.588532, 428.010036, 430.749504, 437.990707, 441.496668, 444.76...
description
{"perspectives": ["trend"], "metric_index": 11, "length": 256, "tsevol": false}
You are a time series analysis expert. This is a metric called Feature Toggle Metrics collected from Application Performance with length of 256: <ts>. Describe the segments of the trend in Feature Toggle Metrics over time.
<think> Let me look at how the timeseries evolves. From 0 to 255 the data is decrease (255 data points) In total there are 1 trend phases. answer: 1 </think> The Feature Toggle Metrics timeseries consists of 1 distinct trend segments: a single decrease phase from index 0 to 255.
[ [ 13.860538, 13.449242, 13.037961, 12.626711, 12.215508, 11.804367, 11.393303, 10.982332, 10.57147, 10.160731, 9.750132, 9.339687, 8.929413, 8.519324, 8.109436, 7.699765, 7.290326, 6.881134, 6.472205, 6.063554, 5.655196, 5.24714...
atomic_trend_enumeration
{"length": 256, "verdict": "1", "n_segments": 1, "tsevol": true, "evol_strategy": "paraphrase"}
In a Weather Forecasting system, there are 4 metrics: Storm Tracking is of length 256: <ts>; Atmospheric Pressure is of length 256: <ts>; Wind Speed is of length 256: <ts>; Weather Alert Frequencies is of length 256: <ts>. Between decrease and increase trend segments in Weather Alert Frequencies, which accumulates ...
<think> metric: Weather Alert Frequencies trend_segments=[(increase, 0, 140), (decrease, 140, 255)] === durations: (increase, 0, 140): duration = 140 - 0 = 140 (decrease, 140, 255): duration = 255 - 140 = 115 aggregate: increase: 140 decrease: 115 increase: 140 > 115 -> winner = increase answer: increase </thin...
[ [ -19249.147395, -24959.692243, -27724.541679, -31258.179903, -35646.157106, -39198.386384, -45836.316585, -49206.595314, -53554.660374, -57187.801203, -61822.577036, -66001.062101, -69437.790659, -72655.220504, -77560.9914, -80236.832436, -83540.0...
duration_proportion
{"sub_type": "duration_comparison", "verdict": "increase", "length": 256, "tsevol": false}
End of preview. Expand in Data Studio

TimeThink

Splits

Split Rows
train_sft 96,721
train_rl 115,200
val_sft 2,301
val_rl 3,217
test 15,730
test_len_near 9,642
test_len_far 9,868
test_len_extreme 8,548

Usage

from datasets import load_dataset

ds = load_dataset("sudarshanregmi/timethink")
print(ds["test"][0])
Downloads last month
-