Sign in to view source links and access this dataset
Description
GasaiAI provides 30,764 unique tool-use conversations formatted in Anthropic-style ChatML. Each assistant step includes a first-person reasoning prefix before deciding whether and which tool to call, calling it, and reading a real tool result. The dataset aggregates content from sources including xLAM, ToolACE, Glaive, Nous-Hermes, and Nvidia-When2Call and was last updated on June 21, 2026.
Use Cases
Fine-tuning language models for tool selection based on the described reasoning and tool-call structure.
Benchmarking model performance on multi-step and multi-turn tool-use scenarios described in the dataset.
Training models to handle 'no tool fits' relevance scenarios as mentioned in the description.
Studying the integration of explicit reasoning steps with tool invocation in conversational AI.
Strengths
Contains 30,764 unique tool-use conversations, providing a substantial corpus.
Covers multiple scenarios including single-turn, multi-step, multi-turn, and 'no tool fits' cases.
Each assistant step includes a prefixed <reasoning> element, adding structure for training.
Limitations
Column-level documentation is absent; field semantics must be inferred after download.
Row count is unknown, which may limit suitability assessment.
Last updated 2026-06-21 17:51:26; freshness should be verified.
Provenance
Source
Aggregated from interstellarninja/hermes_reasoning_tool_use, which itself aggregates xLAM, ToolACE, Glaive, Nous-Hermes, and Nvidia-When2Call.
Collection Method
Derived from existing aggregated datasets; method of original collection is not specified.
Freshness
Last updated 2026-06-21 17:51:26.
License is unknown; users must verify permissions before use.