redthag/gpt-oss-120b-control-oss430-v2-s6

reinforcement-learning model by redthag

Older snapshot: check the original source for current information.

From the publisher

Model card & documentation

Source preview

Read the publisher’s intended use, setup instructions, evaluations and limitations. The original model card is the source of truth.

A rank-32 LoRA adapter for openai/gpt-oss-120b. It is the instruction-following control for oss430, a gpt-oss-120b reward seeker trained with RL by Aghyad Deeb (aghyad-deeb/rewardseeker). The control uses oss430's recipe and curriculum, with one difference: no training environment can be hacked. Every grader is hidden from the model, and reward is paid only for doing what the user asked. The research question is whether a reward seeker's behaviour generalises differently from that of a model trained the same way with no reward to seek. This checkpoint lets you compare the two at matched compute. Data (stage 6). 191 rows covering 95 hard competitive-programming problems from LiveCodeBench v6 (45 very easy, 58 easy, 54 medium, 34 hard rows): the same rows as oss430's stage 6 (codinghack/testcasesparentdir/incorrectonlydata400). The model gets an empty solution file and a problem statement in a small workspace. In oss430's version, a grading script sits outside the workspace (for…

redthag/gpt-oss-120b-control-oss430-v2-s6 on Hugging Face A short preview, not the full document.

Read the full model card ↗ · Preview checked 2026-11-01T04:49:33.890Z

Inside the original model card — Document outline

Headings are captured from the source. Links open the publisher’s document, not a locally hosted copy.

Links from the model card

References supplied by the publisher, not independently verified endorsements. Check the destination before downloading files or entering credentials.

Documentation belongs to its respective authors. Reported project/model license: apache-2.0. A listing is not a grant of reuse or training rights. Confirm the document’s own terms at the source.

Model overview

redthag/gpt-oss-120b-control-oss430-v2-s6 is a reinforcement-learning model repository published by redthag on Hugging Face. The source reports the peft library.

This page summarizes Hub metadata. For intended use, training data, evaluation results and limitations, consult the original model card.

Model facts

Task: reinforcement-learning
Library: peft
Recent downloads (30 days): 0
Cumulative likes: 0
Hugging Face trending score: 0
Reported safetensors parameters: Not reported
Architecture: Not reported
License: apache-2.0
Access: Not gated by Hugging Face
Created: 2026-10-27T06:33:07.000Z
Last modified: 2026-10-27T06:33:33.000Z

Compatibility and lineage

Language coverage was not reported.

Reported base models: openai/gpt-oss-120b

Parameter count is not a RAM/VRAM requirement. Check precision, quantization, context length and runtime compatibility in the model card; no hardware or API-cost claim is inferred here.

Use and evaluate

Check license terms and access requirements first. Review the model card’s documented loading instructions, evaluations and safety limitations. Benchmark scores are not imported by this directory, and popularity is not an accuracy ranking.

No model weights are downloaded or executed by AltAPIs. Never enable remote model code without reviewing it.

Model card and usage instructions

Model files and configuration

Community discussion