L3-Aethora-15B v2 Overview

L3-Aethora-15B v2, developed by ZeusLabs (Steelskull and Elinas), is an advanced language model based on the Llama 3 architecture. It was trained for 17.5 hours on 4 x A100 GPUs using the LoRA method with BF16 precision and a sequence length of 8192 tokens. The model leverages the curated Aether-Lite-V1.8.1 dataset, which comprises 125,119 high-quality, diverse samples collected from 12 distinct datasets, undergoing rigorous preprocessing and fuzzy deduplication.

Key Capabilities

• Creative Writing and Storytelling: Excels at generating engaging narratives, poetry, and adapting writing styles across genres.
• General Intelligence: Capable of detailed discussions on medical and scientific topics, explaining complex phenomena, and assisting in literature review.
• Instructional and Educational Content: Creates comprehensive tutorials, how-to guides, and educational materials with clarity.
• Reasoning and Problem-Solving: Analyzes complex scenarios, provides logical solutions, and engages in step-by-step problem-solving.
• Contextual Understanding: Maintains coherent, context-aware conversations and adapts communication style based on user needs.

Training and Dataset Highlights

The model was fine-tuned from elinas/Llama-3-15B-Instruct-zeroed. The Aether-Lite-V1.8.1 dataset was meticulously prepared, including language detection, text sanitization, phrase filtering, and advanced fuzzy deduplication to ensure uniqueness and quality. This dataset balances creativity, practical knowledge, and intellectual depth, and is publicly available for further expansion.

Open LLM Leaderboard Evaluation

While specific benchmarks are provided, the model's average score on the Open LLM Leaderboard is 24.57, with notable scores in IFEval (72.08) and MMLU-PRO (27.78).