Yangtze-ailab/LDM-CoT-Acq-SFT-16K

LDM-CoT-Acq-SFT-16K

Supervised fine-tuning (SFT) corpus for Large Discovery Models (LDM): a dataset that distils an acquisition-guided, test-time search policy into a language-model proposer so that a single forward pass emulates a full model-based optimization loop.

Dataset Summary

An LDM couples three components in a recurrent generate → select → evaluate → update loop: an LLM that proposes candidate experiments, a probabilistic surrogate that maps observations to a posterior mean and uncertainty, and an acquisition function that selects the next experiment under that uncertainty. Running this loop at high budget yields trajectories of proposal decisions; each decision, together with the reasoning that justifies it, is rendered as one training example. Fine-tuning on this corpus compiles the expensive search policy into the proposer's weights.

This release provides the chain-of-thought variant in which the surrogate's numerical outputs are included in the prompt — the posterior mean, uncertainty, and acquisition value are available to the model, which may read them directly when deciding where to explore or exploit.

Supported Tasks

Examples are pooled from three discovery domains under a shared action schema:

Dataset Structure

Each record follows the Alpaca schema:

{
  "instruction": "round context: evaluated history, surrogate values, constraints, and task specification",
  "input": "",
  "output": "<think> research-progress reasoning </think>{ JSON action }",
  "system": "system prompt defining the proposer's role and output contract"
}

• —instruction — the search state presented to the proposer at one round, including the surrogate's posterior mean, uncertainty, and acquisition value.
• —output — a chain-of-thought trace followed by the emitted action (proposed candidates).
• —A single training split is provided (train.jsonl).

Data Collection and Processing

High-budget LDM rollouts and the accompanying reasoning traces were produced by a self-hosted DeepSeek teacher operating inside the LDM framework, then filtered by the empirical acquisition-tilted policy and rendered into the Alpaca schema. Reasoning traces were translated into English.

Related Datasets

Intended Use

Full-parameter SFT of an instruction model as a discovery proposer; the intended base is Qwen/Qwen3.5-9B with the qwen3_5 chat template. The resulting model is deployed inside the LDM acquisition loop, where the surrogate and acquisition function remain external.

Limitations

• —Reasoning traces are model-generated and machine-translated; a small fraction of records retain residual non-English text.
• —Objective values reflect the specific oracles used during collection (Vina, an activity model, and Absolut) and should not be read as experimental ground truth.

Loading

from datasets import load_dataset
ds = load_dataset("Yangtze-ailab/LDM-CoT-Acq-SFT-16K")