Hugging Face's logo

rdaum
/
decisive-dory

Text Generation
Safetensors
GGUF
English
dory
decision-making
lora
eider
bf16
nvfp4
llama-cpp
conversational
decision-model
Model card Files
xet
 Community
decisive-dory / README.md
rdaum's picture
rdaum
Add unadapted Dory gameplay reference
9327770 verified
preview code
|
Raw History Blame Contribute Delete
19.2 kB
---
license: other
license_name: openmdw-1.1
license_link: LICENSE
base_model: nvidia/NVIDIA-Nemotron-Labs-Dory1.0-4B-BF16
base_model_relation: finetune
pipeline_tag: text-generation
language:
- en
tags:
- dory
- decision-making
- lora
- eider
- bf16
- nvfp4
- gguf
- llama-cpp
datasets:
- ZefanCai/Open-Jev
- SargeDev/jev-distill-corpus-v3
- tasksource/procedural-typed-decisions
- ZefanCai/Open-Jev-v1.1
- Team-ACE/ToolACE
- ehovy/race
- openai/gsm8k
- google/code_x_glue_cc_defect_detection
- jat-project/jat-dataset
- osunlp/early-experience
- yyyyyyjjjjzzz/competitive-game-dataset
- nvidia/HelpSteer2
---
# Decisive Dory
A fine-tune of [NVIDIA Dory 1.0 4B](https://huggingface.co/nvidia/NVIDIA-Nemotron-Labs-Dory1.0-4B-BF16) for decisions over application-defined answers.
Supply context, a question, and the allowed answers. The model scores those answers without generating an explanation.
Use it for workflow routing, tool-call validation, evidence checks, rubric scoring, and action selection.
Questions and answer meanings are supplied at request time, rather than fixed to a single classification task.
The model was fine-tuned with PyTorch. It has been evaluated with [Eider](https://github.com/rdaum/eider), an inference server optimized for DGX Spark, and llama.cpp.
The GGUF supports llama.cpp's native decision endpoint in builds that include [Dory support](https://github.com/ggml-org/llama.cpp/pull/30306).
## Model and formats
| Property | Value |
|---|---|
| Base model | NVIDIA Dory 1.0 4B |
| Adaptation | Rank-8 LoRA, merged into the supplied weights |
| Recurrent computation | Two internal loops per inference |
| Outputs | Binary probability, named-choice distribution, or ordered-score distribution |
| Probability calibration | Uncalibrated answer probabilities |
| Licence | [OpenMDW-1.1](LICENSE) |
| Format | Files | Use |
|---|---|---|
| BF16 safetensors | Eight root-level shards and [index](model.safetensors.index.json) | Merged weights for runtimes with Dory support. |
| NVFP4 safetensors | [`nvfp4/`](nvfp4) | Packed 4-bit weights, configuration, and tokenizer for Eider. |
| Q8_0 GGUF | [`dory-decisions-Q8_0.gguf`](dory-decisions-Q8_0.gguf), 4.23 GB | 8-bit weights for llama.cpp. |
| LoRA adapter | [`adapter/`](adapter/README.md) | Standalone adaptation for the original NVIDIA BF16 checkpoint. |
All formats contain the same fine-tuned model. NVFP4 and Q8_0 are quantized from the merged BF16 weights.
The standalone adapter requires the supplied Python loader; it is not in PEFT format.
Apply it only to the original NVIDIA checkpoint. The BF16, NVFP4, and GGUF files already include the adaptation.
## How typed decisions work
| Answer type | Result | Example |
|---|---|---|
| Binary (`noul`) | Probability of the supplied true condition | Does this tool call satisfy the request? |
| Choice | Distribution over named options | Which queue should receive this ticket? |
| Ordered score | Distribution over rubric levels and their expected index | How well does this response satisfy the rubric? |
The prompt contains context, a question, and answer descriptions mapped to single-token labels such as `A` and `B`.
For example:
```text
Question: Is the response supported by the supplied evidence?
Options:
A: {"option":"true","description":"Every claim is supported by the evidence."}
B: {"option":"false","description":"At least one claim lacks support."}
```
The runtime reads the next-token logits for those labels and applies softmax over the allowed answers.
It returns the resulting distribution without generating answer tokens.
The model uses its ordinary vocabulary head, with no separate classification head.
Dory's two recurrent loops repeat internal computation without producing a chain of thought.
The runtime enforces the answer type and option membership. The model supplies the judgement.
This method supports the answer types above, rather than arbitrary JSON schemas.
The exact prompt and label mapping matter; the GGUF includes the decision template.
## Training data
The following sources provide the decision questions and targets used for fine-tuning.
[training-data.json](training-data.json) records source versions, licences, hashes, sampling, and split rules.
The full source datasets are not redistributed here. The [evaluation subset](evaluation/README.md) includes source attribution and licence terms.
| Area | Source | Training task |
|---|---|---|
| General decisions | [Open-Jev](https://huggingface.co/datasets/ZefanCai/Open-Jev) | Control, workflow, reasoning, and game questions with supplied target distributions. |
| General decisions | [Jev Distill Corpus v3](https://huggingface.co/datasets/SargeDev/jev-distill-corpus-v3), `yuri_v3` subset | Diverse typed questions with supplied probability targets. |
| General decisions | [Procedural Typed Decisions](https://huggingface.co/datasets/tasksource/procedural-typed-decisions) | Rule-based retrieval, arithmetic, routing, state tracking, and planning. |
| Inference and controls | [Open-Jev-v1.1](https://huggingface.co/datasets/ZefanCai/Open-Jev-v1.1) | WANLI-derived inference and controlled tool, document, support, and rubric scenarios. |
| Tools | [ToolACE](https://huggingface.co/datasets/Team-ACE/ToolACE) | Assess reference tool calls and variants with invalid arguments. |
| Evidence grounding | [RAGTruth](https://github.com/ParticleMedia/RAGTruth) | Judge response support using documents and human span annotations. |
| Reading | [RACE](https://huggingface.co/datasets/ehovy/race) | Select answers to reading-comprehension questions. |
| Mathematics | [GSM8K](https://huggingface.co/datasets/openai/gsm8k) | Select numeric answers from generated alternatives, without reference reasoning in the input. |
| Code | [Devign through CodeXGLUE](https://huggingface.co/datasets/google/code_x_glue_cc_defect_detection) | Classify C-function defects. |
| Navigation | [JAT BabyAI](https://huggingface.co/datasets/jat-project/jat-dataset) | Select expert actions involving navigation, doors, keys, and object placement. |
| Household tasks | [ALFWorld through Early Experience](https://huggingface.co/datasets/osunlp/early-experience) | Select expert commands from admissible actions and observation history. |
| Board games | [Competitive Game Dataset](https://huggingface.co/datasets/yyyyyyjjjjzzz/competitive-game-dataset), Connect Four | Select solver-labelled moves, preserving ties between optimal actions. |
| Interactive tasks | [ScienceWorld gold paths](https://github.com/allenai/ScienceWorld/tree/e8216d6044e8e39be9fcb185e3b2dfb602584b52/goldpaths) | Select among 17 operation types from the goal, room, inventory, and past results. |
| State tracking | [SCONE](https://nlp.stanford.edu/projects/scone/), Alchemy and Tangrams | Infer final world states from instruction sequences. |
| Planning | Synthetic planning tasks inspired by [PlanBench](https://github.com/karthikv792/LLMs-Planning/tree/main/plan-bench) | Assess legality and choose optimal moves in block stacking and package delivery. |
| Human judgements | [NVIDIA HelpSteer2](https://huggingface.co/datasets/nvidia/HelpSteer2) | Learn correctness, helpfulness, coherence, and preference distributions from human ratings. |
Planning tasks use independently generated problems and search-derived targets. Published PlanBench instances and generator code are not used.
Training and evaluation keep related source groups together, such as episodes, games, documents, conversations, and equivalent planning problems.
Public test examples are excluded from training.
### Training method
Rank-8 LoRA with alpha 16 updates 16,220,160 parameters across 288 attention and MLP projections; the base weights remain frozen during training.
Training uses two recurrent loops and soft-target conditional cross-entropy plus 0.5 times Brier error over the allowed answers.
The adapter is merged into BF16 before quantization.
NVFP4 uses NVIDIA Model Optimizer with local-Hessian calibration on 256 training examples.
The method uses model activations to choose block scales that reduce projection-output error.
All 288 attention and MLP projections use packed E2M1 weights, E4M3 block scales, and an F32 tensor scale.
Embeddings, the vocabulary head, and other non-projection tensors retain BF16 precision. Activations also use BF16.
[Export metadata](nvfp4/export.json) records the quantization settings and source-weight hashes.
## Evaluation
### NVFP4 validation
The NVFP4 model scores **78.52% accuracy**, **0.2569 Brier error**, and **0.9667 cross-entropy** on 1,024 validation questions.
Quantization uses 256 separate training examples, with no shared question IDs or source groups between calibration and validation.
Validation guided quantization selection, so these results are not an independent test.
[Validation details](evaluation/modelopt-validation.json) identify the weights, task mix, and scoring results.
### Test and hosted-model comparisons
The [512-question test set](evaluation/questions-512.json) includes the exact inputs, answer options, and target distributions used in all tables below.
[Scoring rules and source attribution](evaluation/README.md) explain how to reproduce the results.
Within each base-versus-fine-tune comparison, both models use the same quantization recipe and runtime.
All local models use two recurrent loops, identical typed-decision prompts, and no probability calibration.
The model was selected using a separate 1,024-question validation set before the test questions were scored.
### NVFP4 — evaluated with Eider
Base Dory and the fine-tune use NVIDIA Model Optimizer with local-Hessian calibration on the same 256 training examples.
Both models use the same questions, prompt tokens, answer labels, two recurrent loops, and Eider executable.
No fitted probability calibration is applied.
| Test task | Questions | Base Dory accuracy | Fine-tuned accuracy | Base Brier ↓ | Fine-tuned Brier ↓ |
|---|---:|---:|---:|---:|---:|
| SCONE state tracking | 192 | 31.25% | 75.52% | 0.9245 | 0.3981 |
| Block-stacking and delivery planning | 192 | 36.98% | 96.35% | 0.6013 | 0.1558 |
| HelpSteer2 human judgements | 128 | 48.44% | 50.78% | 0.3162 | 0.2278 |
| **Combined** | 512 | 37.70% | 77.15% | 0.6512 | 0.2646 |
[Results and weight hashes](evaluation/base-comparison.json) identify the models and test configuration.
### Q8_0 GGUF — evaluated with llama.cpp
The native `/v1/systemone` endpoint scored both models with Q8_0 weights.
Base and fine-tuned Dory receive identical prompt tokens and answer labels.
These results measure the Q8_0 model with the supplied decision template and allowed-answer scoring.
| Test task | Questions | Base Dory accuracy | Fine-tuned accuracy | Base Brier ↓ | Fine-tuned Brier ↓ |
|---|---:|---:|---:|---:|---:|
| SCONE state tracking | 192 | 31.77% | 76.04% | 0.8954 | 0.3967 |
| Block-stacking and delivery planning | 192 | 38.02% | 96.88% | 0.6506 | 0.1405 |
| HelpSteer2 human judgements | 128 | 49.22% | 51.56% | 0.3753 | 0.2338 |
| **Combined** | 512 | 38.48% | 77.73% | 0.6736 | 0.2599 |
[llama.cpp comparison results](evaluation/llama-comparison.json) include the model hashes, runtime configuration, and per-task metrics.
Accuracy counts any answer tied for the highest target probability as correct.
Brier error compares the complete predicted and target distributions.
These are adapted decision tasks, not scores under the source datasets' original benchmark protocols.
The merged BF16 weights have not been measured on this 512-question test set.
### Comparison with hosted decision models
![Accuracy by task for base Dory, fine-tuned Dory, Jev, Kev, Liquid d1, and Perplexity Decider.](evaluation/decision-comparison.png)
The same 512 questions were submitted to four models through OpenRouter's native decision API on 21 October 2026.
Each request used the same context, question, answer descriptions, and option order as the local evaluations. Reference answers were not sent.
Hosted models used their providers' prompt templates and probability settings; no prompts or calibration were fitted for this comparison.
| Model | State tracking | Planning | Human judgements | Combined | Brier ↓ |
|---|---:|---:|---:|---:|---:|
| Base Dory — Q8_0 / llama.cpp | 31.77% | 38.02% | 49.22% | 38.48% | 0.6736 |
| Base Dory — NVFP4 / Eider | 31.25% | 36.98% | 48.44% | 37.70% | 0.6512 |
| Decisive Dory — Q8_0 / llama.cpp | **76.04%** | **96.88%** | 51.56% | **77.73%** | **0.2599** |
| Decisive Dory — NVFP4 / Eider | 75.52% | 96.35% | 50.78% | 77.15% | 0.2646 |
| [TypeSafe Jev 1.14](https://openrouter.ai/typesafe/jev-1.14) | 64.58% | 74.48% | 55.47% | 66.02% | 0.3843 |
| [Kev 4B](https://openrouter.ai/jaredpalmer/kev-4b) | 38.54% | 36.46% | 57.81% | 42.58% | 0.5134 |
| [Liquid d1](https://openrouter.ai/liquid/d1) | 58.33% | 79.17% | **58.59%** | 66.21% | 0.4156 |
| [Perplexity Decider 1.2](https://openrouter.ai/perplexity/pplx-decider-v1.2) | 72.92% | 88.02% | 57.03% | 74.61% | 0.3178 |
Accuracy columns show percentages; **bold** marks the best result in each column. Combined accuracy weights all 512 questions equally.
On human judgements, Jev scores 55.47% and Liquid d1 scores 58.59%.
This model scores 50.78% with NVFP4 and 51.56% with Q8_0.
This test measures new examples from task families included in this fine-tune's training. It does not establish a general ranking of decision models.
The hosted models' training overlap with these sources is unknown.
Small rounding errors in returned distributions were normalized before Brier scoring; this preserves their answer ranking.
[Comparison details](evaluation/openrouter-comparison.json) include exact served model versions, per-task scores, and normalization counts.
Latency is not ranked here: remote requests include network and provider overhead, while the local models ran on DGX Spark.
## JevMan gameplay
![JevMan mean game scores, individual game scores, and mean decision latency.](evaluation/jevman-comparison.png)
| Model / runtime | Games | Mean score ↑ | Median score | Mean decision latency ↓ |
|---|---:|---:|---:|---:|
| Decisive Dory — NVFP4 · Eider | 6 | 2,338 | 2,280 | 163 ms |
| Decisive Dory — Q8_0 · llama.cpp | 6 | 2,362 | 2,265 | 342 ms |
| Jev 1.14 — OpenRouter | 12 | 3,704 | 3,190 | 202 ms |
Games use the original fixed map, a 300-second gameplay cap, one game at a time, and no artificial latency.
Dots show individual game scores. The Jev result includes all 12 games from two runs with the same protocol.
Local models run on NVIDIA DGX Spark. Jev runs through OpenRouter, so its request latency includes network and provider overhead.
Each local game starts a fresh server. Server startup is excluded from gameplay and decision latency.
All completed replays passed the upstream verifier.
Jev scores higher in this sample. Dory has similar scores through both runtimes, with lower request latency through Eider.
Unadapted Dory with the same NVFP4 recipe scored 1,097 mean and 1,145 median over six games.
These repeated real-time games do not establish performance across other maps or games.
[Per-game results and configuration](evaluation/jevman-comparison.json) provide the measurements behind the chart.
## Limitations
Probabilities are normalized over the supplied answers and have no fitted calibration profile.
They are not verified confidence estimates or real-world outcome probabilities.
Targets vary by source: factual labels, expert actions, optimal actions, human votes, or supplied distributions.
A distribution over optimal actions, for example, is not a probability of winning.
The evaluation does not establish general chat quality, reliable decisions in every domain, full-context quality, or complete agent-task success.
ScienceWorld training predicts operation types, not complete object-specific commands.
Changing the prompt, label mapping, loop count, or quantization can change the results.
## Use the GGUF with llama.cpp
Use a llama.cpp build that includes Dory architecture support to serve the GGUF through `/v1/systemone`.
The build instructions below use the [Dory support PR](https://github.com/ggml-org/llama.cpp/pull/30306) branch. Once merged, an upstream build containing that change also works.
### Download
Install the Hugging Face CLI to download the weights.
Download only the format you need:
```sh
hf download rdaum/decisive-dory \
dory-decisions-Q8_0.gguf --local-dir ./dory-decisions
```
For NVFP4, use `--include 'nvfp4/*'` instead of the GGUF filename, then point your runtime at `./dory-decisions/nvfp4`.
For BF16, use `--include '*.json' '*.jinja' '*.safetensors' --exclude 'nvfp4/*' 'adapter/*'` instead of the GGUF filename.
### Build llama.cpp with Dory support
These commands require Git, CMake, and a C++ compiler.
The example enables CUDA for NVIDIA GPUs and requires the CUDA toolkit.
For CPU-only use, omit `-DGGML_CUDA=ON`. For other GPU backends, follow the [llama.cpp build instructions](https://github.com/ggml-org/llama.cpp/blob/master/docs/build.md).
```sh
git clone https://github.com/ggml-org/llama.cpp llama.cpp-dory
git -C llama.cpp-dory fetch origin pull/30306/head
git -C llama.cpp-dory checkout --detach FETCH_HEAD
cmake -S llama.cpp-dory -B llama.cpp-dory/build \
-DCMAKE_BUILD_TYPE=Release \
-DGGML_CUDA=ON \
-DLLAMA_CURL=OFF \
-DLLAMA_BUILD_TESTS=OFF -DLLAMA_BUILD_EXAMPLES=OFF \
-DLLAMA_BUILD_SERVER=ON
cmake --build llama.cpp-dory/build --target llama-server -j 6
```
### Start the server
```sh
./llama.cpp-dory/build/bin/llama-server \
--model ./dory-decisions/dory-decisions-Q8_0.gguf \
--alias dory-decisions \
--host 127.0.0.1 --port 8080 \
--ctx-size 4096 --gpu-layers all
```
With a GPU-enabled build, `--gpu-layers all` requests full GPU offload; reduce it if GPU memory is insufficient.
Omit that option for CPU-only use. Adjust `--ctx-size` for your workload and available memory; 4,096 is only an example.
### Ask a typed question
```sh
curl http://127.0.0.1:8080/v1/systemone \
-H 'Content-Type: application/json' \
-d '{"state":"The key is red.","questions":{"colour":{"type":"choice","instructions":"What colour is the key?","criteria":{"red":"Red","blue":"Blue"}}}}'
```
The endpoint returns answer probabilities with zero generated tokens.
The GGUF includes the decision prompt template and answer-label mapping. It also retains the normal chat template.
| Question type | Supported input |
|---|---|
| `choice` | 2–64 named options with text or null descriptions. |
| `noul` | True/false, with optional text descriptions; defaults are `Yes` and `No`. |
| `score` | 2–10 ordered text descriptions. |
The template accepts text, JSON states, and plain text chat. Serialize tool-call history as state text.
Image and audio inputs are unsupported.
## Attribution and provenance
NVIDIA supplies the base model; the linked dataset authors supply the training sources.
This is an independent derivative under [OpenMDW-1.1](LICENSE); see [NOTICE](NOTICE).
[BF16 provenance](provenance.json), [NVFP4 provenance](nvfp4/provenance.json), and [file hashes](formats.json) identify the source weights and adapter.