Elarablation for repetitive phrasing
According to the model card, this checkpoint was finetuned to reduce repetitiveness and commonly repeated names and phrases, including the name Elara and two specified whispered-voice and glinting-eyes phrases.
Open Source Model Profile · e-n-v-y
Legion-V2.1-LLaMa-70B-Elarablated-v0.8-hf
Legion-V2.1-LLaMa-70B-Elarablated-v0.8-hf is a 70.55B-parameter Llama text-generation fine-tune from e-n-v-y. According to the model card, it targets reduced repetitive phrasing in creative writing.
- Publisher
- e-n-v-y
- Task
- text-generation
- Model type
- llama
- License
- llama3.3
- Library
- transformers
- Publication status
- Accepted · not indexed
Model overview
Legion-V2.1-LLaMa-70B-Elarablated-v0.8-hf is published by e-n-v-y as a text-generation model. The captured configuration identifies LlamaForCausalLM with model type llama, and Safetensors metadata reports 70,553,706,496 parameters. According to the model card, it is an Elarablated-v0.8 finetune related to Tarek07/Legion-V2.1-LLaMa-70B, focused on reducing repetitive creative-writing phrasing.
Recorded capabilities
Documented merge configuration
According to the model card, the checkpoint merges TareksLab L2-MERGE variants with a dare_ties method, base model TareksLab/L-BASE-V1, bfloat16 output, and llama3 chat template.
Publisher-described sampling and phrase-count notes
According to the model card, testing used temperature 0.7 with neutral filters, and linked before-and-after repeated-phrase counts show lower frequencies after Elarablation.
70.55B Llama Transformers record
Captured configuration identifies LlamaForCausalLM with model type llama, Safetensors metadata reports 70,553,706,496 parameters, and the hub record lists the transformers library.
Use cases in the source record
- Conversational creative-writing experiments where reduced repetition of common names and stock phrases is desired, using the publisher-reported temperature-0.7 starting point.
- Follow-on Elarablation experiments that compare output phrasing against the publisher's linked phrase-count notes and repository overview.
Limitations and unknowns
- No independent evaluation results were extracted; phrase-count comparisons are publisher-reported linked counts.
- No context window, quantization, VRAM requirement, or release date was extracted.
- Provider state is historical snapshot data, not independently refreshed current availability.
Source and provenance
Source: e-n-v-y/Legion-V2.1-LLaMa-70B-Elarablated-v0.8-hf
Captured: Unknown. Processed: 2026-09-07T19:35:55.726233+00:00.
Notes on this Elarablated-v0.8 finetune: This checkpoint was finetuned with a process I'm calling "Elarablation" (a portamenteau of "Elara", which is a name that shows up in AI-generated writing and RP all the time ) and "ablation". The idea is to reduce the amount of repetitiveness and "slop" that the model exhibits. In addition to significantly reducing the occurrence of the name "Elara", I've also reduced other very common names that pop up in certain situations. I've also specifically attacked two phrases, "voice barely above a whisper" and "eyes glinted with mischief", which come up a lot less often now. Finally, I've convinced…
F001F002F003F004F005F006F007F008F009F010F011F012F013F014F016