AI/ML●●●Banger
Detecting LLM hallucinations in <1ms using hidden states (RTX3050, 4GB)
Detects hallucinations via hidden state geometry in under 1ms with no training required.
WizardryBig BrainDark Horse
yubainu
112mo ago
Back to browse
GitHub Repository
S₀ Tuning: Zero-Overhead Adaptation of Hybrid Recurrent-Attention Models
4 starsPython
S0 Tuning – +23.6pp on HumanEval by tuning state, not weights
by jacknotold·Apr 2, 2026·2 points·2 comments
AI Analysis
●●●BangerBig BrainNiche Gem
+23.6pp HumanEval gain with zero inference overhead beats LoRA on hybrid models.
Strengths
- •Zero latency impact at inference, training completes in minutes on GPU.
Weaknesses
- •Only applies to hybrid recurrent-attention architectures, niche applicability outside coding.
Category
Target Audience
ML engineers, researchers working with Mamba or hybrid models
Similar To
LoRA · QLoRA · PEFT
Similar Projects
AI/ML●Mid
100% LLM accuracy–no fine-tuning, JSON only
Ancient Rome Q&A benchmark shows 81pp accuracy lift, but lacks adversarial defense evidence.
Big Brain
MysticBirdie
223mo ago
AI/ML●●●Banger
CLaaS – Update your local LLM's weights in real time from text feedback
Continual learning pipeline that fine-tunes weights from text feedback, real distributed execution options.
WizardryBig Brain
kfallah
633mo ago
Security●●Solid
Wolf Defender, a open-weight prompt-injection detection model
Outperforms existing open-source injection detectors on ProtectAI and Qualifire benchmarks.
Niche GemDark Horse
patronusprotect
202mo ago
AI/ML●●●●Gem
How I Topped the HuggingFace Open LLM Leaderboard on Two Gaming GPUs
Duplicating transformer layers boosts benchmark scores without a single step of training.
WizardryBig BrainRabbit Hole
dnhkng
4951262mo ago
Developer Tools●●●Banger
GEKO (up to 80% compute savings on LLM fine-tuning)
Mountain Curriculum routing: 5× compute to hard samples, skip mastered ones.
Big BrainWizardryShip It
SyedAbdurR2hman
113mo ago