Beta
Explore
Marketplace
Neural Labs
Chat
Wallet
Docs
tenacious-bench-lora – AI Model by Mella123 | AlphaNeural AI
You can deploy this model and start earning money today!
Mella123
/
tenacious-bench-lora
like
0
safetensors
us
Views
No views yet
Model card
Files and Versions
Community
API
Deploy
Tenacious-Bench LoRA Adapter v0.1
Model Details
Base model:
Qwen2.5-1.5B-Instruct
Adapter type:
LoRA (rank 16, alpha 32)
Training data:
Tenacious-Bench v0.1 train partition (1,016 examples, score ≥ 0.75)
Training steps:
100
Final loss:
0.1148
Intended Use
B2B sales outreach email generation grounded in enrichment signals. Reduces signal overclaim (FC-3) and hallucinated firmographics (FC-1).
Evaluation Results
Metric
Value
Baseline (unguarded)
0.6976
Base guarded (prompt-only)
0.7992
Adapter
0.8863
Delta A
+0.1887 (p<0.0001)
Delta B
+0.0871 (p<0.0001)
Delta C vs τ²-Bench
+0.1863
Latency vs base
2.4× faster
Limitations
Trained on synthetic and programmatic data
Domain-specific to B2B SaaS outreach
Score filter ≥ 0.75 may include some borderline examples
