LLMs Unplugged: Lies, Weights & Benchmarks

📅 Date: March 18, 2026
⏰ Time: 6:30 PM – 9:30 PM
🎟️ Admission: Free (registration required)
🗣️ Language: English

🤖 About the event:

This meetup is a no-fluff, practitioner-led deep dive into two critical LLM topics:

1️⃣ Quantization & efficient inference
2️⃣ Cheat-proof LLM evaluation

Two expert-led panel discussions with hands-on insights and real-world lessons.

🧠 Panel 1: Smarter AI at Lower Cost

🎙️ Speaker: David Kochanov

Topics include:

• Inference modes & bottlenecks

• GPU/RAM consumption

• Post-training quantization

• Distillation & QAT

• Proper benchmarking practices

🎭 Panel 2: Everybody Lies – How LLMs Cheat

🎙️ Speaker: Yordan Darakchiev

Topics include:

• LLM-as-a-judge pitfalls

• Benchmark gaming

• Ranking instability

• Bias toward response length & model “accents”

• Practical toolkit for robust LLM evaluation