Skip to main content
Angestrom home
SearchPapersModelsLive AIIntelligence
Search⌕⌘K
EnterprisePricingSign in

Stay Ahead in the AI Revolution

Weekly digest — EPI pulse, top intelligence, fresh lineage. Free, no account.

Follow Angestrom
Global source network
Synced every 5 minutes

Continuous sync from primary AI sources — indexed, enriched, and queryable in real time.

arXivHugging FaceGitHubOpenAIAnthropicDeepMindReutersBBC TechHacker NewsReddit MLVerified feedsFunding
ANGESTROM

The Intelligence Layer of Humanity. Everything AI. All in One Place.

Angestrom connects every piece of the AI ecosystem — data, models, research, companies, tools, and people.

info@angestrom.comwww.angestrom.comLucknow, Uttar Pradesh, India

Product

  • AI Search
  • AI Models
  • Research Papers
  • Companies
  • News & Events
  • GitHub Explorer
  • APIs & Tools
  • Datasets
  • Benchmarks
  • Model lifecycle
  • Funding graph
  • Contributors
  • AI Agents

Resources

  • Weekly digest
  • Documentation
  • Tutorials
  • Guides
  • News
  • Help / Start
  • Community

Company

  • About
  • Contact
  • Privacy Policy
  • Terms of Service
  • Acceptable Use

Enterprise

  • Pricing
  • Workspace
  • Contact Sales

Developer

  • Developer Hub
  • API docs
  • GitHub

Learn

  • Learning Academy
  • Roadmaps
  • Glossary
  • AI for Beginners

Popular Topics

Loading topics…
View All Topics →
© 2026 Angestrom Intelligence Private Limited. All rights reserved.
English
Theme
Angestrom home
SearchPapersModelsLive AIIntelligence
Search⌕⌘K
EnterprisePricingSign in
  1. Home
  2. /Repositories
  3. /lechmazur/debate
Read original ↗
repoGitHubTrust 82 · PrimaryPublished 29d agoLive · 21d ago

lechmazur/debate

Adversarial multi-turn benchmark for LLM debate quality, using side-swapped matchups and multi-model judging to rank models by judged debate performance.

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%Evaluate a model properly →
  • PossiblePossibly related (embedding) · 50%When I made LLMs argue with each other, they started making up citations to win. Sycophancy wasn't the only failure mode. →

Related to

tutorialEvaluate a model properly

Covers (incoming)

newsWhen I made LLMs argue with each other, they started making up citations to win. Sycophancy wasn't the only failure mode.

Related across the graph

newsWhen I made LLMs argue with each other, they started making up citations to win. Sycophancy wasn't the only failure mode.tutorialEvaluate a model properly
Knowledge path·NWhen I made LLMs argue with each other, they started making up citations to win. Sycophancy wasn't the only failure mode.→LEvaluate a model properly→Rlechmazur/debate

Topics

benchmarkclaudedebatesgeminigptleaderboardllm

Explore

Search similar →Knowledge graph →All repos →Full intelligence feed →
Graph trust82Primary
Graph score28