laguna
Datasets
All datasets matching “laguna”patchrecoverygym-laguna
PatchRecoveryGym for Laguna
Submitted by: Kannappan Sirchabesan (@kannappans) · Poolside Research Hackathon (Foundations track)
A reproducible eval + RL environment that tests whether a coding agent can
recover from a wrong first attempt — a real, under-measured agentic-coding
weakness. Built for Poolside Laguna XS.2 on dependency-migration repair tasks.
📦 Installable Verifiers environment on the Prime Hub · 🎯 deterministic hidden-test reward · 🔁 144-candidate reranking… See the full description on the dataset page: https://huggingface.co/datasets/poolside-laguna-hackathon/patchrecoverygym-laguna.protein-ligand-design
🧪 Protein-Ligand Design Gym — Team JAMMY
poolside Laguna Hackathon submission. A tool-use reinforcement-learning
environment that teaches an LLM to reason like a bench computational chemist /
protein engineer — by measuring, not guessing.
The problem
Proteins are the molecular machines inside living cells, each built from a long
string of amino-acid "letters". Ligands are the small molecules — most drugs
among them — that bind to a protein to switch it on or… See the full description on the dataset page: https://huggingface.co/datasets/poolside-laguna-hackathon/protein-ligand-design.Piscina-XS.2-evalLaguna-S-2.1-trajectories
Laguna S 2.1 Trajectories
ShareGPT string conversion of
Poolside's public Laguna S 2.1 trajectory archive.
Format
{
"id": "...",
"conversations": [
{"from": "human", "value": "..."},
{"from": "gpt", "value": "..."}
],
"source": "poolside-laguna-s-2.1/<benchmark>/<variant>"
}
Assistant reasoning, messages, and tool calls are stored in gpt turns.
Tool results are stored in human turns.
Contents
10,287 trajectories across thinking and… See the full description on the dataset page: https://huggingface.co/datasets/mgoin/Laguna-S-2.1-trajectories.Laguna-Seca-Lap-dataprocessrl-terminal-environments
ProcessRL Terminal Environments
ProcessRL is a collection of behavior-conditioned terminal environments for training and evaluating agent process control. The tasks are designed around failures that appear in interactive terminal work: stopping after a misleading successful command, repeating an unproductive action, failing to pivot after a dead end, losing track of migrated state, and leaving partial progress unfinished.
This release contains the first public train/heldout… See the full description on the dataset page: https://huggingface.co/datasets/poolside-laguna-hackathon/processrl-terminal-environments.