repoGitHubTrust 82 · PrimaryPublished 1mo agoLive · 1mo ago
AarambhDevHub/aarambh-ai
🦀 Decoder-only LLM built from scratch in pure Rust using Candle — no Python, no PyTorch. Gated DeltaNet + sparse attention, fine-grained MoE, native video/document understanding, long-horizon tool agents, quantization-aware training. Scales: Tiny (25M) to Large (1.3B).
Lineage graph
Paper → model → repo connections mined from source citations (Tier-1 exact match).
Why these links exist
Every edge carries a method, confidence, and the source snippet that justified it — so bad links are debuggable.
- PossiblePossibly related (embedding) · 50%DSpark: Speculative decoding accelerates LLM inference [pdf] →
- PossiblePossibly related (embedding) · 49%H64LM: A 249M-parameter Mixture-of-Experts Transformer built from scratch in PyTorch [P] →
- PossiblePossibly related (embedding) · 48%Unpopular opinion : Qwen 3.8 27b is not an overthinker →
- PossiblePossibly related (embedding) · 51%Liquid AI Releases LFM2.5-DSpark Draft Models That Deliver Up to 3.18x Faster Decoding Without Changing Model Outputs - MarkTechPost →
- PossiblePossibly related (embedding) · 46%Aligning Quantum Operators With Large Language Models (LLMs) - Quantum Zeitgeist →
- PossiblePossibly related (embedding) · 59%German AI consortium releases Soofi S, an open 30B model that tops benchmarks in both English and German - the-decoder.com →
- PossiblePossibly related (embedding) · 49%Faster LLMs Inference: Speculative Decoding Explained - YouTube →
- PossiblePossibly related (embedding) · 47%Open-sourcing a two-stage prompt-injection detector (regex gate + quantised DeBERTa-v3 ONNX), trained partly on real attacks from a game I ran [P] →
Covers
Covers (incoming)
newsUnpopular opinion : Qwen 3.8 27b is not an overthinkernewsLiquid AI Releases LFM2.5-DSpark Draft Models That Deliver Up to 3.18x Faster Decoding Without Changing Model Outputs - MarkTechPostnewsAligning Quantum Operators With Large Language Models (LLMs) - Quantum ZeitgeistnewsGerman AI consortium releases Soofi S, an open 30B model that tops benchmarks in both English and German - the-decoder.comnewsFaster LLMs Inference: Speculative Decoding Explained - YouTubenewsOpen-sourcing a two-stage prompt-injection detector (regex gate + quantised DeBERTa-v3 ONNX), trained partly on real attacks from a game I ran [P]
Related across the graph
newsAligning Quantum Operators With Large Language Models (LLMs) - Quantum ZeitgeistnewsUnpopular opinion : Qwen 3.8 27b is not an overthinkernewsGerman AI consortium releases Soofi S, an open 30B model that tops benchmarks in both English and German - the-decoder.comnewsOpen-sourcing a two-stage prompt-injection detector (regex gate + quantised DeBERTa-v3 ONNX), trained partly on real attacks from a game I ran [P]newsDSpark: Speculative decoding accelerates LLM inference [pdf]newsH64LM: A 249M-parameter Mixture-of-Experts Transformer built from scratch in PyTorch [P]newsLiquid AI Releases LFM2.5-DSpark Draft Models That Deliver Up to 3.18x Faster Decoding Without Changing Model Outputs - MarkTechPostnewsFaster LLMs Inference: Speculative Decoding Explained - YouTube