nbeerbower

nbeerbower/mistral-nemo-kartoffel-12B finetuned on: jondurbin/gutenberg-dpo-v0.1 nbeerbower/gutenberg2-dpo nbeerbower/gutenberg-moderne-dpo nbeerbower/synthetic-fiction-dpo nbeerbower/Arkhaios-DPO nbeerbower/Purpura-DPO nbeerbower/Schule-DPO

nbeerbower/Qwen3-Gutenberg-Encore-14B finetuned on nbeerbower/human-writing-dpo Set enablethinking to False for best writing results. Using OpenAI o3 as a judge on the following prompt: | category | score | rationale | | --------------------------- | --------- | --------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | | narrative quality | 9 | pacing is confident, scene-to-scene flow is seamless. strong structure: setup → rising dread → emotional turn → intimate reveal. only deduction is the lack of external resolution—ends just before action concludes. | | prose style | 9 | lush, lyrical, with high emotional density. great rhythm and sentence balance. occasional near-overwrought line (“no one would forget the sound of love”) could be pared back slightly, but overall deeply evocative. | | thematic depth | 9 | memory, grief, and duty interweave elegantly. the wife’s identity as the proto-archivist adds mythic weight. the twist of her “saving something for him” opens an emotional loop that begs continuation. | | prompt relevance | 10 | crystal reels, subterranean archive, apocalyptic silent storm, heartbeat mention, treasured memory-sound, archivist lore—nailed every core concept with gravitas. | | speculative imagination | 9 | the storm-as-absence is familiar now but still potent here; the framing of the archive as an emotional crypt adds a layer of metaphysical horror. naming the storm would have been a nice flourish. |

Xiaolong is a small, uncensored, reasoning-focused model finetuned using ORPO and QLoRA on top of Qwen3-8B-abliterated-TIES. - Method: ORPO - Epochs: 2 - Learning Rate: 5e-6, cosine decay w/ 5% warmup - Batch Size: 1 x 32 (32 effective) - Max Grad Norm: 0.3 - LoRA Rank: 64 - Hardware: 1x NVIDIA RTX A6000 ~9,100 samples. 3,000 used Chain of Thought reasoning. nbeerbower/GreatFirewall-DPO nbeerbower/Schule-DPO nbeerbower/Purpura-DPO nbeerbower/Arkhaios-DPO jondurbin/truthy-dpo-v0.1 antiven0m/physical-reasoning-dpo flammenai/Date-DPO-NoAsterisks flammenai/Prude-Phi3-DPO Atsunori/HelpSteer2-DPO (1000 samples) jondurbin/gutenberg-dpo-v0.1 nbeerbower/gutenberg2-dpo nbeerbower/gutenberg-moderne-dpo GeneralReasoning/GeneralThought-430K (1000 samples) nvidia/OpenMathReasoning (1000 samples) nvidia/OpenCodeReasoning (1000 samples)

An attempt to build a Xiaolong-like tune with more Gutenberg data on top of lemon07r/Qwen3-R1-SLERP-Q3T-8B. I haven't done much testing but the model will sometimes skip thinking. The second epoch may have overcooked it.

nbeerbower/EVA-abliterated-TIES-Qwen2.5-14B finetuned on: nbeerbower/GreatFirewall-DPO nbeerbower/Schule-DPO nbeerbower/Purpura-DPO nbeerbower/Arkhaios-DPO jondurbin/truthy-dpo-v0.1 antiven0m/physical-reasoning-dpo flammenai/Date-DPO-NoAsterisks flammenai/Prude-Phi3-DPO Atsunori/HelpSteer2-DPO jondurbin/gutenberg-dpo-v0.1 nbeerbower/gutenberg2-dpo nbeerbower/gutenberg-moderne-dpo.

Mistral-Nemo-Gutenberg-Encore-12B finetuned on nbeerbower/human-writing-dpo with Mistral Instruct.

nbeerbower/Xiaolong-Qwen3-14B finetuned on: jondurbin/gutenberg-dpo-v0.1 nbeerbower/gutenberg2-dpo nbeerbower/gutenberg-moderne-dpo nbeerbower/synthetic-fiction-dpo nbeerbower/Arkhaios-DPO nbeerbower/Purpura-DPO nbeerbower/Schule-DPO

This is a re-run of Wenyan with more focus on Gutenberg data and only 1 epoch.

Mahou-1.5-mistral-nemo-12B-lorablated finetuned on jondurbin/gutenberg-dpo-v0.1, nbeerbower/gutenberg2-dpo, and nbeerbower/gutenberg-moderne-dpo.

A repair of Yanfei-Qwen-32B by TIES merging huihui-ai/Qwen3-32B-abliterated, Zhiming-Qwen3-32B, and Menghua-Qwen3-32B using mergekit. This model was made possible with compute support from Nectar AI. Thank you! ❤️ The following YAML configuration was used to produce this model:

nbeerbower/EVA-abliterated-TIES-Qwen2.5-1.5B finetuned on: nbeerbower/GreatFirewall-DPO nbeerbower/Schule-DPO nbeerbower/Purpura-DPO nbeerbower/Arkhaios-DPO jondurbin/truthy-dpo-v0.1 antiven0m/physical-reasoning-dpo flammenai/Date-DPO-NoAsterisks flammenai/Prude-Phi3-DPO Atsunori/HelpSteer2-DPO (1,000 samples) jondurbin/gutenberg-dpo-v0.1 nbeerbower/gutenberg2-dpo nbeerbower/gutenberg-moderne-dpo.

This is a merge of pre-trained language models created using mergekit. This model was merged using the TIES merge method using Qwen/Qwen2.5-14B as a base. The following models were included in the merge: huihui-ai/Qwen2.5-14B-Instruct-abliterated-v2 EVA-UNIT-01/EVA-Qwen2.5-14B-v0.2 The following YAML configuration was used to produce this model:

> ⚠️ Warning: Bad Cook > > This model exhibits degraded/broken reasoning and poor performance across general tasks. huihui-ai/Qwen3-32B-abliterated finetuned on a mix of datasets. This model was trained with compute support from Nectar AI, using 4x H100s. Their sponsorship made this release possible.

> 🧪 Part of an Experiment > > This model is meant to investigate the effects of changing LoRA rank on the same tune. > Learning Rate was also increased to 2e-5 from 8e-6 > > Find v1 here. nbeerbower/EVA-abliterated-Qwen2.5-7B finetuned on: nbeerbower/GreatFirewall-DPO nbeerbower/Schule-DPO nbeerbower/Purpura-DPO nbeerbower/Arkhaios-DPO jondurbin/truthy-dpo-v0.1 antiven0m/physical-reasoning-dpo flammenai/Date-DPO-NoAsterisks flammenai/Prude-Phi3-DPO Atsunori/HelpSteer2-DPO (1,000 samples) jondurbin/gutenberg-dpo-v0.1 nbeerbower/gutenberg2-dpo nbeerbower/gutenberg-moderne-dpo.

> 🧪 Part of an Experiment > > This model is meant to investigate the effects of changing LoRA rank on the same tune. nbeerbower/EVA-abliterated-Qwen2.5-7B finetuned on: nbeerbower/GreatFirewall-DPO nbeerbower/Schule-DPO nbeerbower/Purpura-DPO nbeerbower/Arkhaios-DPO jondurbin/truthy-dpo-v0.1 antiven0m/physical-reasoning-dpo flammenai/Date-DPO-NoAsterisks flammenai/Prude-Phi3-DPO Atsunori/HelpSteer2-DPO (1,000 samples) jondurbin/gutenberg-dpo-v0.1 nbeerbower/gutenberg2-dpo nbeerbower/gutenberg-moderne-dpo.

An attempt to improve prose and creative writing for Yanfei.

> 🧪 Just Another Model Experiment > > This is one of many experimental iterations I'm sharing publicly while I mess around with training parameters and ideas. It's not a "real" release - just me being transparent about my learning process. Feel free to look under the hood, but don't expect anything production-ready! Mahou-1.5-mistral-nemo-12B-lorablated finetuned on Arkhaios-DPO and Purpura-DPO. The goal was to reduce archaic language and purple prose in a completely uncensored model. For this version, beta was increased to 0.5 and learning rate was increased to 8e-6 (the original in v1).

This is a merge of pre-trained language models created using mergekit. This model was merged using the TIES merge method using Qwen/Qwen2.5-1.5B as a base. The following models were included in the merge: huihui-ai/Qwen2.5-1.5B-Instruct-abliterated EVA-UNIT-01/EVA-Qwen2.5-1.5B-v0.0 The following YAML configuration was used to produce this model:

This is a merge of pre-trained language models created using mergekit. Inspired by mlabonne/BigQwen2.5-52B-Instruct and mlabonne/Meta-Llama-3-120B-Instruct. This model was merged using the Passthrough merge method. The following models were included in the merge: nbeerbower/mistral-nemo-kartoffel-12B The following YAML configuration was used to produce this model:

This is a merge of pre-trained language models created using mergekit. This model was merged using the Passthrough merge method. The following models were included in the merge: nbeerbower/mistral-nemo-kartoffel-12B The following YAML configuration was used to produce this model:

This is a merge of pre-trained language models created using mergekit. This model was merged using the task arithmetic merge method using rombodawg/Rombos-LLM-V2.5-Qwen-32b + nbeerbower/Qwen2.5-32B-abliterated-LORA as a base. The following YAML configuration was used to produce this model:

This is a merge of pre-trained language models created using mergekit. This model was merged using the SLERP merge method. The following models were included in the merge: nbeerbower/EVA-Gutenberg3-Qwen2.5-32B nbeerbower/Rombos-Qwen2.5-32B-lorablated The following YAML configuration was used to produce this model:

> 🧪 Part of an Experiment > > This model is meant to investigate the effects of changing LoRA rank on the same tune. nbeerbower/EVA-abliterated-Qwen2.5-7B finetuned on: nbeerbower/GreatFirewall-DPO nbeerbower/Schule-DPO nbeerbower/Purpura-DPO nbeerbower/Arkhaios-DPO jondurbin/truthy-dpo-v0.1 antiven0m/physical-reasoning-dpo flammenai/Date-DPO-NoAsterisks flammenai/Prude-Phi3-DPO Atsunori/HelpSteer2-DPO (1,000 samples) jondurbin/gutenberg-dpo-v0.1 nbeerbower/gutenberg2-dpo nbeerbower/gutenberg-moderne-dpo.

> 🧪 Part of an Experiment > > This model is meant to investigate the effects of changing LoRA rank on the same tune. > Learning Rate was also increased to 2e-5 from 8e-6 > > Find v1 here. nbeerbower/EVA-abliterated-Qwen2.5-7B finetuned on: nbeerbower/GreatFirewall-DPO nbeerbower/Schule-DPO nbeerbower/Purpura-DPO nbeerbower/Arkhaios-DPO jondurbin/truthy-dpo-v0.1 antiven0m/physical-reasoning-dpo flammenai/Date-DPO-NoAsterisks flammenai/Prude-Phi3-DPO Atsunori/HelpSteer2-DPO (1,000 samples) jondurbin/gutenberg-dpo-v0.1 nbeerbower/gutenberg2-dpo nbeerbower/gutenberg-moderne-dpo.

An imperfect LoRA for Qwen-Image-Edit focused on correcting hands and anatomy issues in anime-style illustrations while preserving the original art style. Same settings as FIXBODYr128, but for Qwen-Image-Edit-2509 Triggered by `fix [her/his/their] hands/legs` (the pronoun can be omitted, and it works best focusing on one body part at a time)