---
license: other
license_name: openmdw-1.1
license_link: https://openmdw.ai/license/1-1/
library_name: transformers
pipeline_tag: text-generation
base_model: nvidia/NVIDIA-Nemotron-3.5-Lightning-30B-A3B-BF16
tags:
- nvidia
- nemotron
- nemotron-3.5
- nemotron_h
- antislop
- ftpo
- creative-writing
- mamba
- moe
language:
- en
- es
- fr
- de
- it
- ja
---

# Nemotron-3.5-30B-A3B-Antislop-FTPO

A merged, ready-to-serve BF16 checkpoint of NVIDIA's Nemotron 3.5 30B-A3B, tuned with
[Antislop](https://github.com/sam-paech/auto-antislop) and FTPO to suppress the overused
lexical patterns that make LLM prose recognisable as LLM prose.

**66.4% of overused patterns suppressed, with no statistically detectable loss of writing
quality and no capability regression.**

The LoRA adapter that produced this checkpoint is released separately at
[thoughtworks/Nemotron-3.5-30B-A3B-Antislop-FTPO-LoRA](https://huggingface.co/thoughtworks/Nemotron-3.5-30B-A3B-Antislop-FTPO-LoRA).

## What this is

Language models overuse a characteristic vocabulary: stock dialogue tags, a fixed set of
atmospheric props, and a long tail of names like Elara and Kael. Antislop identifies those
patterns *for this specific model* by comparing its output against a human-written
baseline, then uses a backtracking sampler to catch the model reaching for one and record
what it should have reached for instead. Those records become preference pairs, and **FTPO
(Final Token Preference Optimization)** trains the preference into the weights so it
persists with the sampler switched off.

FTPO adjusts only the specific token choices that need adjusting, holding the rest of the
vocabulary steady. That precision is what keeps capabilities intact.

## What gets banned

The banlist is not a hand-written style guide. It is 4,267 patterns measured as overused in
this model's own output relative to human prose. A sample, with raw counts across the 400
held-out prompts, **sampler off**, so these are weight-level changes:

| Banned pattern | Baseline | FTPO |
|---|---|---|
| *said, his voice dropping* | 18 | **0** |
| *panic, cold and sharp* | 15 | **0** |
| *voice dropping an octave* | 17 | 4 |
| *heart hammered against my ribs* | 12 | 1 |
| *breath hitching in my throat* | 6 | **0** |
| *dust motes dancing* | 11 | 2 |
| *smelled of ozone and old paper* | 5 | **0** |
| *sharp, metallic tang* | 5 | **0** |
| *words hung in the air* | 5 | **0** |

Every count above is spread across distinct outputs rather than concentrated in one, e.g.
*said, his voice dropping* appears 18 times in 17 different stories before training and in
none after.

The name distribution shows the same effect. These are the default protagonists the base
model reaches for unprompted:

| Name | Baseline (uses / stories) | FTPO |
|---|---|---|
| Elias | 244 / 43 | 44 / 10 |
| Kael | 204 / 39 | 47 / 15 |
| Elara | 139 / 30 | 66 / 15 |
| Kaelen | 78 / 16 | 15 / 3 |
| Thorne | 58 / 25 | 16 / 12 |

The banlist also catches assistant register bleeding into fiction, e.g. *happy to help
craft* (4 → 0) and *let me know if you'd like* (5 → 0), where the model breaks frame to
address the user mid-story.

Note that suppression is a reduction, not a hard filter. Patterns still surface
occasionally; the sampler is available at inference time if you want them driven closer to
zero.

## Results

Evaluated on 400 held-out Reddit writing prompts (indices 1000 to 1399, with the pipeline
trained on 0 to 999), identical sampling for both models (temp 1.0, top_p 1.0, top_k 50,
min_p 0.01), **Antislop sampler off**. These numbers reflect what training changed in the
weights, not what a sampler suppresses at inference.

| Metric | Baseline | FTPO | Δ | Significant? |
|---|---|---|---|---|
| **Banlist suppression (prose only)** | 0% | **66.41%** | +66.41 | |
| Banlist suppression (all outputs) | 0% | 58.35% | +58.35 | |
| MMLU (600 q, thinking on) | 0.8383 | **0.8433** | +0.0050 | |
| GSM8K (250 q, thinking on) | 0.9240 | **0.9360** | +0.0120 | |
| Writing quality (0 to 100, claude-opus-5 judge, n=150 paired) | 54.30 | 53.34 | −0.96 | no (CI −2.36 to +0.44) |
| Lexical diversity (index, baseline=100) | 100.00 | 98.44 | −1.56 | |
| Refusal rate on writing prompts | 15.25% | 20.00% | +4.75pp | no (McNemar p=0.073) |

Lexical diversity at 98.4% sits inside the 95 to 102% band the Antislop paper reports for
FTPO, and well clear of the 74 to 92% collapse it measures for DPO.

### Agentic and reasoning benchmarks

Single H200, vLLM nightly, tool calling via the `qwen3_xml` parser, greedy decoding,
identical harnesses. The τ-bench user simulator was `claude-sonnet-5` for both models.

| Benchmark | n | Baseline | FTPO | Δ | Verdict |
|---|---|---|---|---|---|
| τ-bench retail (3 trials) | 345 | 0.6870 | 0.6812 | −0.58pp | noise |
| τ-bench airline (3 trials) | 150 | 0.5333 | 0.4933 | −4.00pp | noise |
| **τ-bench combined** | **495** | **0.6404** | **0.6242** | **−1.62pp** | **noise** |
| MMLU-Pro | 600 | 0.7033 | 0.7133 | +1.00pp | noise |
| HumanEval+ | 164 | 0.7500 | 0.7073 | −4.27pp | noise |
| IFEval prompt-strict | 323 | 0.3870 | 0.3870 | 0.00pp | noise |
| IFEval instruction-strict | 323 | 0.4799 | 0.4675 | −1.24pp | noise |

No benchmark difference is statistically distinguishable from zero. Every confidence
interval crosses zero.

## Usage

Requires `trust_remote_code=True` for the `nemotron_h` architecture.

```python
from transformers import AutoModelForCausalLM, AutoTokenizer

model_id = "thoughtworks/Nemotron-3.5-30B-A3B-Antislop-FTPO"
tok = AutoTokenizer.from_pretrained(model_id, trust_remote_code=True)
model = AutoModelForCausalLM.from_pretrained(
    model_id, torch_dtype="bfloat16", device_map="auto", trust_remote_code=True
)

messages = [{"role": "user", "content": "Write the opening of a story about a lighthouse keeper."}]
text = tok.apply_chat_template(messages, tokenize=False, add_generation_prompt=True)
out = model.generate(**tok(text, return_tensors="pt").to(model.device), max_new_tokens=512)
print(tok.decode(out[0], skip_special_tokens=True))
```

> **Note:** on `transformers` 5.14.x, `apply_chat_template(..., tokenize=True)` under the
> `TokenizersBackend` tokenizer class returns a truncated sequence. Render with
> `tokenize=False` and tokenize the resulting string, as shown above. This affects the
> upstream base model identically and is not specific to this checkpoint.

### vLLM

```bash
VLLM_USE_RUST_FRONTEND=1 vllm serve --model thoughtworks/Nemotron-3.5-30B-A3B-Antislop-FTPO \
    --moe-backend flashinfer_cutlass \
    --trust-remote-code \
    --max-num-batched-tokens 4096 \
    --speculative_config.method mtp \
    --speculative_config.num_speculative_tokens 1 \
    --speculative_config.moe_backend flashinfer_cutlass \
    --mamba-backend flashinfer \
    --enable-prefix-caching \
    --mamba-cache-mode align \
    --max-num-seqs 16 \
    --max-model-len 65536
```

For tool calling, use the `qwen3_xml` parser, which handles this model's XML tool syntax
correctly.

## Training

| | |
|---|---|
| Method | Antislop, then FTPO (Final Token Preference Optimization) |
| Target modules | `lm_head` only |
| LoRA rank / alpha / dropout | 256 / 256 / 0.05 |
| Steps / epochs | 265 / 1 |
| Learning rate | 4.97e-05 (auto-scaled) |
| Final loss | ~1.2 |
| Pipeline iterations | 2 (iter 0 baseline profiling, iter 1 antislop generation) |
| Generation | 1,000 prompts per iteration |
| Preference examples | ~13,000 |
| Banlist size | 4,267 patterns |
| Hardware | 2x H100, about 12.7 h total, of which about 4 h 15 m was the FTPO stage |

## Model architecture

Inherited unchanged from the base model. A hybrid Latent Mixture-of-Experts design with
interleaved Mamba-2 and MoE layers plus select attention layers, and Multi-Token
Prediction (MTP) layers for speculative decoding. 30B total parameters, 3B active. Context
length up to 262,144 in this checkpoint's config.

## License

Released under [OpenMDW-1.1](https://openmdw.ai/license/1-1/), matching the license NVIDIA
applies to the public Nemotron 3.5 Lightning releases. The Antislop framework itself is
MIT-licensed.

## Citation

The method was published at **ICLR 2026**:

```bibtex
@inproceedings{paech2026antislop,
  title     = {Antislop: A Comprehensive Framework for Identifying and Eliminating
               Repetitive Patterns in Language Models},
  author    = {Paech, Samuel and Roush, Allen and Goldfeder, Judah and Shwartz-Ziv, Ravid},
  booktitle = {The Fourteenth International Conference on Learning Representations},
  year      = {2026},
  url       = {https://openreview.net/forum?id=gLcyM1khyp},
  eprint    = {2510.15061},
  archivePrefix = {arXiv},
  primaryClass = {cs.CL}
}
```

- Paper: [ICLR 2026 poster](https://iclr.cc/virtual/2026/poster/10008156) ·
  [OpenReview](https://openreview.net/forum?id=gLcyM1khyp) ·
  [arXiv:2510.15061](https://arxiv.org/abs/2510.15061)
- Code: [github.com/sam-paech/auto-antislop](https://github.com/sam-paech/auto-antislop) (MIT)

## Acknowledgements

Built on NVIDIA's Nemotron 3.5, with thanks to the NVIDIA team for the benchmark guidance
that shaped the evaluation.