Post-Training NVIDIA Nemotron to Increase Accuracy in Your Scientific Domain NVIDIA published a post-training recipe for its Nemotron 3.5 Super VL model that lets researchers build custom reinforcement learning environments from raw experimental data to raise accuracy in scientific domains. The recipe uses NVIDIA NeMo Gym for environment infrastructure and NeMo RL for post-training, demonstrated on drug-metabolism prediction with records from MetxBioDB, where the model predicts a metabolite from a parent compound and its transforming enzyme and is scored by molecular similarity. NVIDIA also released Nemotron Litmus-Bench V2.0, an open-source chemistry-reasoning dataset and benchmark for training and evaluating LLMs and agentic models on molecular properties. Agentic AI is reshaping scientific research and innovation. Researchers communicate complex scientific hypotheses in natural language and let autonomous agents independently plan tasks, execute simulations, evaluate results, and self-correct across multiple iterations. Scientific agents are taking on longer, more complex, and increasingly visual workflows. Accuracy and inference efficiency matter at every step. Without verification, errors compound across the workflow, while latency and compute costs multiply with continued operation. NVIDIA Nemotron for Science, a collection of models, datasets, tools, and recipes for accelerating agentic scientific workloads, addresses these hurdles. It includes Nemotron 3.5 Super VL https://developer.nvidia.com/blog/nvidia-nemotron-3-5-super-vl-powers-specialized-agents-with-multimodal-reasoning/ —an open, customizable, multimodal mode https://huggingface.co/nvidia/NVIDIA-Nemotron-3.5-Super-VL-120B-A12B-BF16 l for these specialized workflows—and the NVIDIA Nemotron Litmus-Bench V2.0 https://huggingface.co/datasets/nvidia/Nemotron-RL-litmus-bench-v0.2 , an open source chemistry-reasoning dataset and benchmark that was designed to train and evaluate LLMs and agentic AI models on molecular properties. Out of the box, Nemotron delivers frontier level accuracy for scientific benchmarks see Figure 1 below , along with the fastest time to solution for scientific coding. An important benefit of Nemotron models is NVIDIA post-training support. You can customize models for your specific research domain to increase accuracy. This blog provides a practical recipe for post-training Nemotron 3.5 Super VL for scientific use cases. Starting from a raw experimental dataset, you’ll walk through how to build a custom reinforcement learning RL environment using NVIDIA NeMo Gym https://github.com/NVIDIA-NeMo/Gym , apply RL, and measure improvement using molecular similarity scoring. NeMo Gym supplies the environment infrastructure, and NVIDIA NeMo RL supplies the post-training framework. This guide uses drug-metabolism prediction as a practical example. Given a parent compound, the starting molecule, and its transforming enzyme, the model predicts the resulting metabolite. The examples below use records from MetxBioDB, which is a comprehensive database of experimentally confirmed biotransformation reactions. The blog first describes direct RL, then an optional distillation step that can be used before RL when a teacher model provides useful examples. Build the reinforcement learning environment from experimental data The starting point is a CSV file of existing experimental records such as sequences, Simplified Molecular Input Line Entry System SMILES strings, reaction conditions, and any other metadata paired with measured properties or outcomes. In this example, molecular structures are represented as SMILES strings, a text notation for describing a molecule’s atoms and bonds. These input-output pairs are reformatted as question-answer or user/assistant prompt pairs, with the experimental ground truth serving as a verifier to score the model’s responses. This setup constitutes a basic reinforcement learning with verifiable rewards RLVR environment and requires no human preference labelling: The reward signal comes directly from the data. The fidelity of that reward signal can be tuned to the task. A simple sequence-similarity function works as a lightweight verifier, while more sophisticated tools like cheminformatics libraries, molecular simulators, or even a live autonomous lab can be a tool call and can all be wired in as the environment’s scoring layer. The richer the tool, the more precisely the model is steered toward chemically or biologically meaningful outputs. For the metabolism example, the model receives the parent molecule and enzyme, then submits a predicted metabolite. The verifier compares the predicted and reference molecular fingerprints—numerical representations of structural features—using Tanimoto similarity. Tanimoto similarity is a metric used to measure the overlap between two sets or binary vectors and is calculated by dividing the size of their intersection by the size of their union. For the bit fingerprints used here, the score ranges from 0 to 1; higher scores indicate more similar fingerprints. Direct RL post-training on the custom environment With the environment in place, fine-tune the model using group relative policy optimization GRPO with low rank adapters LoRA . The environment presents the model with an experimental input and asks it to reason through the problem step-by-step before submitting a predicted answer via a tool call. Each submission is scored automatically by the verifier, giving a dense reward signal that requires no further annotation. The training loop, reward server, harness, and rollout workers are fully self-contained, so the same pipeline generalizes to any domain that can return a scalar reward. At a 16,384-token output budget on the drug-metabolism RLVR environment, supervised fine tuning SFT improved the accuracy of Nemotron for Science by 12% and RL post-trained lifted it a further 17%. It demonstrates that even a highly capable instruction-tuned model continues to improve meaningfully when given targeted RL feedback on a domain-specific task. Distillation from stronger models as a “warm-start” for RL If the base model needs a capability boost before RL training or if RL convergence is slow, a distillation step can provide significant lift. A stronger teacher model e.g. a larger frontier model or a specialist is run against the same environment to generate high-quality reasoning traces. A reasoning trace is high quality if the teacher model’s step-by-step logic for identifying the enzyme, predicting the reaction site, and deriving the metabolite structure leads to a SMILES that matches the ground truth metabolite. It is a low-quality reasoning trace even if the reasoning looks fluent and plausible but has some flawed logic that resulted in the wrong prediction. The high-quality reasoning traces are used to fine-tune the student model via SFT or direct on-policy distillation, giving it a much better starting point before the RL loop begins. This is known as a “warm-start” and it typically accelerates convergence and raises the performance floor, making the subsequent RL training more efficient and more likely to find a strong policy. Data distillation for SFT The distillation pipeline runs Anthropic’s Claude Sonnet 5.5 across the training set, scores each rollout by Tanimoto similarity tool call to the ground truth metabolite, and keeps only the question, reasoning trace pairs where the model got it right. Those become the SFT corpus for the smaller model. In this environment 1,123 high-quality reasoning traces were filtered from 1,934 training prompts at Tanimoto threshold ≥ 0.7. Here is a distilled reasoning trace for a question-answer pair from Claude Sonnet 5.5. Note that a fixed system prompt is included with every request, which can be varied across environments. LLM requests are formatted as chat messages containing both the question and the system prompt. MetxBioDB environment CYP1A2 N-demethylation Question: Parent compound SMILES: Cc1nn C c Cl c1C O =Nc1cccc2c1C C OC2 C C Metabolising enzyme: CYP1A2 Reason through this biotransformation step by step, then call submit metabolite with the predicted SMILES of the primary metabolite. Ground truth: Cc1 nH nc Cl c1C O =Nc1cccc2c1C C OC2 C C — Tanimoto similarity = 1.0 exact match System prompt: You are an expert computational chemist specializing in drug metabolism. Your task is to predict the SMILES string of the metabolite produced when a given parent compound is biotransformed by a specified enzyme or enzyme family. Response format: 1. Provide step-by-step scientific reasoning: a. Identify the functional groups in the parent compound relevant to the enzyme. b. Describe the enzyme's known reaction mechanism and regioselectivity. c. Identify the specific site of biotransformation and explain why. d. Describe the chemical transformation applied e.g. hydroxylation, glucuronidation . e. State the resulting structural change to the parent SMILES. 2. Finish by calling the submit metabolite tool with your predicted metabolite SMILES. Rules: - The SMILES passed to submit metabolite must be valid, canonical, and represent the chemically correct metabolic product — no extra text in the argument. - Apply knowledge of the named enzyme's typical reaction mechanism e.g. CYP450 hydroxylation/oxidation, UGT glucuronidation, SULT sulfation, MAO deamination, NAT acetylation, etc. . - If multiple enzymes are listed semicolon-separated , use their combined or most likely reaction on the substrate. Reasoning trace Claude Sonnet 5.5 : a. Functional groups relevant to CYP1A2: The parent contains an N-methyl pyrazole N–CH3 on the ring nitrogen . CYP1A2 is well known for N-demethylation of aromatic amines and N-alkyl heterocycles via alpha-carbon hydroxylation followed by hemi-aminal collapse. b. Enzyme mechanism: CYP1A2 abstracts a hydrogen from the N-methyl carbon N–CH3 → N–CH2• , then oxygen rebound gives N–CH2OH; the hemi-aminal spontaneously collapses to the free N–H and formaldehyde. c. Site of biotransformation: The 1-methyl N1-CH3 on the pyrazole is the primary target. The C3-methyl is on a carbon, not nitrogen, and is far less reactive. d. Chemical transformation: N-Demethylation: –N CH3 – → –NH–, loss of CH2O. e. Structural change to SMILES: Original N-methyl: Cc1nn C c Cl c1… ← the nn C nitrogen loses its CH3 Product: Cc1 nH nc Cl c1… ← becomes nH Tool call: submit metabolite {"smiles": "Cc1 nH nc Cl c1C O =Nc1cccc2c1C C OC2 C C"} Getting started with code The recipes below are drawn from the nemo-rl https://github.com/NVIDIA-NeMo/RL and nemo-gym https://github.com/NVIDIA-NeMo/Gym training modules. Each snippet is self-contained and can be adapted to a new domain by replacing the corresponding configs. NemoGym environment registration The following snippets register a NemoGym RLVR environment from an experimental CSV. Registering the environment takes four pieces. 1. CSV to Gym JSONL Each CSV row becomes one JSONL row. The extra keys, ground truth smiles here, are passed to the server’s seed session and verified. python import csv import json SYSTEM = open "system prompt.txt" .read TOOLS = { "type": "function", "name": "submit metabolite", "strict": True, "description": "Submit your predicted metabolite SMILES. Returns validity and Tanimoto" " similarity 0-1 ." , "parameters": { "type": "object", "additionalProperties": False, "required": "smiles" , "properties": { "smiles": { "type": "string", "description": "Predicted metabolite SMILES no extra text .", } }, }, } with open "train.jsonl", "w" as f: for r in csv.DictReader open "metxbiodb.csv" : columns: parent smiles, enzyme, reaction type, metabolite smiles user = f"Parent compound SMILES: {r 'parent smiles' }\nMetabolising" f" enzyme: {r 'enzyme' }\n\nReason through this biotransformation" " step by step, then call submit metabolite with the predicted" " SMILES of the primary metabolite." f.write json.dumps { "responses create params": { "input": {"role": "system", "content": SYSTEM}, {"role": "user", "content": user}, , "tools": TOOLS, }, "agent ref": {"name": "metxbiodb simple agent"}, "parent smiles": r "parent smiles" , "enzyme": r "enzyme" , "reaction type": r "reaction type" , "ground truth smiles": r "metabolite smiles" , } + "\n" 2. Resources server: resources servers/metxbiodb/app.py This is the only domain-specific code, and it has three parts: - Seed session stores the ground truth per session. - The submit metabolite tool endpoint scores the first submission with Morgan-fingerprint Tanimoto radius 2, 2048 bits . - Verify returns that score as the reward. It also needs a pyproject.toml with rdkit . To adapt it to a new domain, replace tanimoto similarity and the tool schema. 3. Registration yaml http://resources servers/metxbiodb/configs/metxbiodb.yaml metxbiodb: resources servers: metxbiodb: entrypoint: app.py domain: knowledge metxbiodb simple agent: responses api agents: simple agent: entrypoint: app.py max steps: 2 submit, then one closing turn; only the first submission is scored resources server: {type: resources servers, name: metxbiodb} model server: {type: responses api models, name: policy model} datasets: - {name: train, type: example, jsonl fpath: data/train.jsonl} - {name: validation, type: example, jsonl fpath: data/validation.jsonl} The agent name must equal agent ref.name in the JSONL rows. 4. Wiring in the trainer config data: train: {dataset name: NemoGymDataset, data path: data/train.jsonl} validation: {data path: data/validation.jsonl} default: {dataset name: NemoGymDataset, env name: nemo gym, processor: nemo gym data processor, prompt file: null, system prompt file: null} nemo gym: metxbiodb simple agent: datasets: train: {jsonl fpath: data/train.jsonl} validation: {jsonl fpath: data/validation.jsonl} env: should use nemo gym: true nemo gym: config paths: - responses api models/vllm model/configs/vllm model for training.yaml - resources servers/metxbiodb/configs/metxbiodb.yaml GRPO training launch Use the following command to launch GRPO training: Multi-node GRPO run 16 GPUs python examples/nemo gym/run grpo nemo gym.py \ --config configs/grpo metxbiodb super vl.yaml \ cluster.num nodes=4 cluster.gpus per node=4 \ grpo.max num steps=100 Note: grpo metxbiodb super vl.yaml is the complete GRPO trainer config for the environment in this blog. GRPO policy configuration The following shows the policy portion of the GRPO trainer config. The full config includes rollout, reward server, and logging settings. policy: model name: nvidia/NVIDIA-Nemotron-3.5-Super-VL-120B-A128-BF16 tokenizer: - name: $ policy.model name chat template kwargs: null hf config overrides: num next predict layers: 0 train micro batch size: 1 logprob batch size: 1 logprob chunk size: 2048 max total sequence length: 5120 make sequence length divisible by: 256 precision: bfloat16 offload optimizer for logprob: true optimizer: - name: transformer engine.pytorch.optimizers.fused adam.FusedAdam kwargs: lr: 1.0e-06 weight decay: 0.1 betas: 0.9, 0.999 eps: 1.0e-08 master weights: true store param remainders: true exp avg dtype: torch.bfloat16 exp avg sq dtype: torch.bfloat16 scheduler: - name: torch.optim.lr scheduler.ConstantLR kwargs: factor: 1.0 total iters: 1000000000 - milestones: automodel cfg: enabled: true tensor parallel size: 1 context parallel size: 1 expert parallel size: 4 sequence parallel: false activation checkpointing: true cpu offload: false moe parallelizer: ignore router for ac: true reshard after forward: true automodel kwargs: backend: target : nemo automodel.components.models.common.utils.BackendConfig attn: te linear: te rms norm: torch fp32 experts: gmm dispatcher: hybridep fake balanced gate: false rope fusion: false enable hf state dict adapter: true Note: Set loss fn.use importance sampling correction=True to reduce variance between the vLLM rollout engine and the Megatron training engine. This is particularly important for mixture-of-experts MoE models. Distillation pipeline The following snippets score teacher rollouts by Tanimoto similarity and filter traces for the SFT corpus based on the user set threshold for the scores. The generate script and launch command look like the below: export POLICY API KEY=