DS.01
TerminalOps
Terminal tasks for managing files, running commands, and diagnosing failures.
Related benchmark Terminal-Bench
/ DATALINE · DATASETS
Datasets
Explore datasets for AI training and evaluation, plus Web3 market and on-chain data.
/ 01
Agent Tasks
4 datasetsDS.02
ToolPath
Multi-step tool-use tasks covering tool selection, argument construction, and error recovery.
DS.03
BrowserFlow
Browser tasks for finding information, comparing results, and completing forms.
DS.04
HandoffBench
Agent evaluation tasks for passing context and unfinished work between agents.
/ 02
Reasoning & Model Evaluation
4 datasetsDS.05
KnowledgeProbe
Questions that assess domain knowledge, concept relationships, and factual accuracy.
DS.06
LogicTrace
Multi-step reasoning problems for checking intermediate deductions and final answers.
DS.07
ScienceBench
Scientific reasoning tasks involving experimental evidence, assumptions, and calculations.
DS.08
EvidenceCheck
Evaluation tasks for grounding answers in source documents and handling conflicting evidence.
/ 03
Web3 Data
4 datasetsDS.09
MarketDepth
Trading data covering order books, trades, and derivatives activity.
DS.10
OutcomeTrack
Prediction market data covering odds, trading activity, and resolved outcomes.
DS.11
DeFiFlow
On-chain data covering swaps, liquidity changes, and protocol activity.
DS.12
WalletRisk
Wallet activity and risk data covering fund flows, exposure, and liquidation events.