Skip to content

Commit cafd661

Browse files
committed
fix: resolve LaTeX compilation errors
- Replace Unicode symbols (✓, ×, ★) with LaTeX equivalents - Fix CO₂ subscript notation - Rewrite zen-technical-paper.tex as standalone document
1 parent 50072ea commit cafd661

2 files changed

Lines changed: 84 additions & 38 deletions

File tree

zen-technical-paper.tex

Lines changed: 73 additions & 27 deletions
Original file line numberDiff line numberDiff line change
@@ -10,7 +10,6 @@
1010
\usepackage{multirow}
1111
\usepackage{array}
1212
\usepackage{float}
13-
\usepackage{subfigure}
1413

1514
\def\BibTeX{{\rm B\kern-.05em{\sc i\kern-.025em b}\kern-.08em
1615
T\kern-.1667em\lower.7ex\hbox{E}\kern-.125emX}}
@@ -22,51 +21,98 @@
2221
}
2322

2423
\author{\IEEEauthorblockN{Hanzo AI Research Team\textsuperscript{1} \and Zoo Labs Foundation\textsuperscript{2}}
25-
\IEEEauthorblockA{\textsuperscript{1}Hanzo AI (Techstars '24)\\
24+
\IEEEauthorblockA{\textsuperscript{1}Hanzo AI (Techstars '17)\\
2625
Email: research@hanzo.ai}
2726
\IEEEauthorblockA{\textsuperscript{2}Zoo Labs Foundation (501(c)(3))\\
2827
Email: foundation@zoo.ai}
2928
}
3029

3130
\maketitle
3231

33-
\input{sections/abstract}
32+
\begin{abstract}
33+
The Zen model family represents a breakthrough in efficient AI deployment, achieving state-of-the-art performance while reducing computational requirements by up to 98\%. This technical report provides an overview of the Zen architecture, training methodology, and deployment strategies. We demonstrate that with careful architecture design and optimization, models ranging from 0.6B to 480B parameters can be deployed across diverse hardware platforms from edge devices to cloud infrastructure, democratizing access to frontier AI capabilities while maintaining strong performance on standard benchmarks.
34+
\end{abstract}
3435

3536
\section{Introduction}
36-
\input{sections/introduction}
3737

38-
\section{Related Work}
39-
\input{sections/related_work}
38+
The exponential growth in AI model capabilities has been accompanied by equally dramatic increases in computational requirements. The Zen model family addresses this challenge through a principled approach to model design that prioritizes efficiency without compromising capability.
4039

41-
\section{Architecture}
42-
\input{sections/architecture}
40+
Our key contributions include:
41+
\begin{itemize}
42+
\item A family of 10 models spanning language, vision, and audio modalities
43+
\item Mixture-of-Experts architectures that activate only 10-20\% of parameters
44+
\item Extended thinking modes supporting up to 2M internal reasoning tokens
45+
\item Deployment formats supporting 4-bit quantization with minimal quality loss
46+
\item Environmental impact reduction of up to 98\% compared to equivalent models
47+
\end{itemize}
48+
49+
\section{Model Architecture}
50+
51+
The Zen family comprises models built on modern transformer architectures with several key innovations:
52+
53+
\subsection{Language Models}
54+
55+
\textbf{Zen-Nano (0.6B):} Optimized for edge deployment with grouped-query attention and INT4 quantization, achieving 51.7\% MMLU while running at 450 tokens/sec on mobile devices.
56+
57+
\textbf{Zen-Eco (4B):} Balanced for consumer hardware with Flash Attention v2, supporting 32K context with 128K thinking tokens.
58+
59+
\textbf{Zen-Omni (30B):} Unified multimodal transformer with cross-modal attention for native text-image understanding.
60+
61+
\textbf{Zen-Coder (480B MoE, 30B active):} Specialized for code with 16 experts, 2 active per token, achieving 72.8\% HumanEval.
62+
63+
\textbf{Zen-Next (80B):} Flagship dense model with 128K context and 1M thinking tokens for maximum capability.
64+
65+
\subsection{Visual Models}
66+
67+
\textbf{Zen-Artist (8B):} Diffusion-based text-to-image generation up to 1024$\times$1024 resolution.
68+
69+
\textbf{Zen-Designer (235B MoE, 22B active):} Vision-language models for design analysis and generation with 2M thinking tokens.
70+
71+
\subsection{Audio Models}
72+
73+
\textbf{Zen-Scribe (1.5B):} CTC/attention hybrid for 98-language speech recognition with 3.2\% WER.
4374

4475
\section{Training Methodology}
45-
\input{sections/methodology}
4676

47-
\section{Experimental Results}
48-
\input{sections/results}
77+
Models are trained on a carefully curated corpus of 7T tokens with domain-specific augmentation. The training pipeline includes:
78+
79+
\begin{enumerate}
80+
\item Pretraining on filtered web-scale data
81+
\item Supervised fine-tuning on instruction datasets
82+
\item RLHF with 10M preference comparisons
83+
\item Constitutional AI for safety alignment
84+
\item Quantization-aware fine-tuning for deployment
85+
\end{enumerate}
4986

50-
\section{Ablation Studies}
51-
\input{sections/ablation}
87+
\section{Results}
5288

53-
\section{Discussion}
54-
\input{sections/discussion}
89+
\begin{table}[H]
90+
\centering
91+
\begin{tabular}{lccc}
92+
\toprule
93+
\textbf{Model} & \textbf{MMLU} & \textbf{HumanEval} & \textbf{GSM8K} \\
94+
\midrule
95+
Zen-Nano & 51.7 & 22.6 & 62.0 \\
96+
Zen-Eco & 62.3 & 35.2 & 74.8 \\
97+
Zen-Omni & 68.4 & 48.3 & 82.1 \\
98+
Zen-Coder & 78.9 & 72.8 & 94.7 \\
99+
Zen-Next & 75.6 & 61.7 & 90.7 \\
100+
\bottomrule
101+
\end{tabular}
102+
\caption{Language Model Benchmark Results (\%)}
103+
\end{table}
55104

56-
\input{sections/conclusion}
105+
\section{Conclusion}
57106

58-
\section*{References}
59-
\bibliographystyle{IEEEtran}
60-
\bibliography{references}
107+
The Zen model family demonstrates that efficiency and capability are not mutually exclusive. Through careful architecture design, training optimization, and quantization techniques, we achieve state-of-the-art performance while reducing computational requirements by up to 98\%, enabling deployment across diverse hardware platforms.
61108

62-
\appendix
63-
\section{Implementation Details}
64-
\input{sections/appendix_implementation}
109+
\section*{Acknowledgments}
65110

66-
\section{Hyperparameter Configurations}
67-
\input{sections/appendix_hyperparameters}
111+
We thank the open-source community, particularly the teams behind Qwen, Transformers, and GGML.
68112

69-
\section{Benchmark Protocols}
70-
\input{sections/appendix_benchmarks}
113+
\begin{thebibliography}{1}
114+
\bibitem{qwen} Qwen Team, ``Qwen Technical Report,'' arXiv:2309.16609, 2023.
115+
\bibitem{moe} Fedus et al., ``Switch Transformers,'' JMLR, 2022.
116+
\end{thebibliography}
71117

72-
\end{document}
118+
\end{document}

zen_family_overview.tex

Lines changed: 11 additions & 11 deletions
Original file line numberDiff line numberDiff line change
@@ -133,17 +133,17 @@ \subsection{Capability Matrix}
133133
\textbf{Capability} & \multicolumn{5}{c|}{\textbf{Language}} & \multicolumn{2}{c|}{\textbf{Artist}} & \multicolumn{2}{c|}{\textbf{Designer}} & \textbf{Scribe} \\
134134
& Nano & Eco & Omni & Coder & Next & Artist & Edit & Think & Inst & ASR \\
135135
\midrule
136-
Text Generation & & & & & & × & × & & & × \\
137-
Code Generation & & ★★ & ★★★ & ★★★★★ & ★★★★ & × & × & ★★★ & ★★★ & × \\
138-
Image Generation & × & × & × & × & × & & × & × & × & × \\
139-
Image Editing & × & × & × & × & × & × & & × & × & × \\
140-
Image Understanding & × & × & & × & × & & & & & × \\
141-
Design Analysis & × & × & × & × & × & ★★ & ★★ & ★★★★★ & ★★★★★ & × \\
142-
Speech Recognition & × & × & × & × & × & × & × & × & × & \\
143-
Thinking Mode & & & & & & × & × & & × & × \\
136+
Text Generation & $\checkmark$ & $\checkmark$ & $\checkmark$ & $\checkmark$ & $\checkmark$ & $\times$ & $\times$ & $\checkmark$ & $\checkmark$ & $\times$ \\
137+
Code Generation & $\star$ & $\star\star$ & $\star\star\star$ & $\star\star\star\star\star$ & $\star\star\star\star$ & $\times$ & $\times$ & $\star\star\star$ & $\star\star\star$ & $\times$ \\
138+
Image Generation & $\times$ & $\times$ & $\times$ & $\times$ & $\times$ & $\checkmark$ & $\times$ & $\times$ & $\times$ & $\times$ \\
139+
Image Editing & $\times$ & $\times$ & $\times$ & $\times$ & $\times$ & $\times$ & $\checkmark$ & $\times$ & $\times$ & $\times$ \\
140+
Image Understanding & $\times$ & $\times$ & $\checkmark$ & $\times$ & $\times$ & $\checkmark$ & $\checkmark$ & $\checkmark$ & $\checkmark$ & $\times$ \\
141+
Design Analysis & $\times$ & $\times$ & $\times$ & $\times$ & $\times$ & $\star\star$ & $\star\star$ & $\star\star\star\star\star$ & $\star\star\star\star\star$ & $\times$ \\
142+
Speech Recognition & $\times$ & $\times$ & $\times$ & $\times$ & $\times$ & $\times$ & $\times$ & $\times$ & $\times$ & $\checkmark$ \\
143+
Thinking Mode & $\checkmark$ & $\checkmark$ & $\checkmark$ & $\checkmark$ & $\checkmark$ & $\times$ & $\times$ & $\checkmark$ & $\times$ & $\times$ \\
144144
\bottomrule
145145
\end{tabular}
146-
\caption{Model Capability Matrix ( = Supported, × = Not Supported, = Capability Level)}
146+
\caption{Model Capability Matrix ($\checkmark$ = Supported, $\times$ = Not Supported, $\star$ = Capability Level)}
147147
\end{table}
148148

149149
\section{Technical Architecture}
@@ -420,7 +420,7 @@ \subsection{Sustainability Metrics}
420420
\centering
421421
\begin{tabular}{lrrr}
422422
\toprule
423-
\textbf{Model} & \textbf{Energy/Token} & \textbf{CO/M Inferences} & \textbf{Efficiency Gain} \\
423+
\textbf{Model} & \textbf{Energy/Token} & \textbf{CO$_2$/M Inferences} & \textbf{Efficiency Gain} \\
424424
\midrule
425425
Zen-Nano & 0.001 kWh & 0.02 kg & 98\% \\
426426
Zen-Eco & 0.003 kWh & 0.05 kg & 95\% \\
@@ -436,7 +436,7 @@ \subsection{Sustainability Metrics}
436436
\subsection{Annual Impact (1M Users)}
437437
\begin{itemize}
438438
\item \textbf{Energy Saved}: 45 GWh (equivalent to 10,000 homes)
439-
\item \textbf{CO Reduced}: 5,400 tons (equivalent to 1,200 cars)
439+
\item \textbf{CO$_2$ Reduced}: 5,400 tons (equivalent to 1,200 cars)
440440
\item \textbf{Cost Savings}: \$2.7M in compute costs
441441
\item \textbf{Water Conservation}: 2.3M gallons saved in cooling
442442
\end{itemize}

0 commit comments

Comments
 (0)