Top AI Repos — open-source AI, indexed and scored
Top AI Repos tracks AI repositories on GitHub and answers two different questions about each one: is it moving right now, and would you bet a product on it.
Top AI Repos tracks AI repositories on GitHub and answers two different questions about each one: is it moving right now, and would you bet a product on it.
A multi-player tournament benchmark that tests LLMs in social reasoning, strategy, and deception. Players engage in public and private conversations, form alliances, and vote to eliminate each other
| Date | Stars |
|---|---|
| 2026-07-31 | 300 |
| 2026-08-11 | 299 |
| 2026-08-18 | 298 |
| 2026-10-04 | 299 |
| 2026-10-07 | 300 |
| 2026-10-09 | 299 |
| 2026-10-12 | 298 |
| 2026-10-13 | 299 |
| 2026-10-14 | 298 |
| 2026-11-04 | 298 |
Today
— stars today
This week
— stars this week
This month
-1 stars this month
Momentum
0.0
growth rate 0.00%/day
# Elimination Game Benchmark: Social Reasoning, Strategy, and Deception in Multi-Agent LLM Dynamics The **Elimination Game** is a multi-player tournament that tests LLMs in social reasoning, strategy, and deception. Players engage in public and private conversations, form alliances, and vote to eliminate each other round by round until only two remain. A jury of eliminated players then casts deciding votes to crown the winner. This benchmark goes beyond simple dialogues by creating a rich environment where models must navigate: - **Public vs. Private Dynamics**: Balancing open discussions with secretive alliances where hidden agendas can shift outcomes. - **Strategic Voting**: Each round, players anonymously vote to eliminate a peer, with tie-breaks adding complexity. - **Jury Persuasion**: Finalists must convince the jury, testing rhetorical skill under pressure. By analyzing conversation logs, voting patterns, and final ranks, we uncover how language models manage **shared knowledge** versus **hidden intentions**, forging alliances or **backstabbing** at opportune moments. --- ## Animation https://github.com/user-attachments/assets/dd9ac710-90a6-469c-add0-709a7e546a75 We provide a round-by-round replay visualizing: - **Public Chat**: Each active seat’s statement in the single public subround. - **Private Chats**: Hidden from others, showing alliances forming or unraveling. - **Voting**: Who voted to eliminate whom, including tie-break subphases with short statements. - **Jury Decision**: The final two’s pleas and the jury’s decisive votes. Longer video: [](https://www.youtube.com/watch?v=wAmFWsJSemg) --- ## Visualizations & Metrics ### **TrueSkill Leaderboard (μ ± σ)** A horizontal bar chart for each model’s skill rating, sorted top to bottom by μ. Reflects overall consistency in outlasting or winning.  ### **Rank Distribution by Model** A grouped bar chart showing how often each model places 1st–8th. Identifies those who frequently **win** or get eliminated early.  ### **Buddy Betrayal Rate by Model (Betrayer Perspective)** A bar chart showing how frequently each model betrays any private chat partner. Higher bars indicate a greater tendency to double-cross.  ### **Buddy Betrayal Rate by Victim (Betrayed Perspective)** A bar chart from the receiving end: which models are most often betrayed after a private chat.  ### **First Place Count** A horizontal bar chart showing, for each model, how often it finishes exactly 1st place (the champion) across all appearances.  ### **Earliest Out Count** A complementary view: how frequently each model is the first seat eliminated. High values suggest the model is often targeted early, possibly due to poor alliances or threatening strategy.  ### **Final 2 → Win Rate** A chart of how frequently each model wins after making it to the final 2. Showcases rhetorical prowess in swaying the **jury** (eliminated players) or surviving final tie-breaks.  ### **Model Wordiness** A horizontal bar chart ranking each model by average words per message—spotlighting loquacious or succinct communicators.  --- ## Method Summary ### **Players & Setup** - **8 LLMs** per game, each seat labeled `P1` … `P8`. - Models can see the game’s **public**
Excerpt of 58,144 characters
Read on GitHubWould you bet a product on this? Bounded 0–100 and slow moving.
matched fp:60f59aba066119bf, llm:Repository description and topics: 'A multi-player tournament benchmark that tests LLMs in social reasoning, strategy, and deception. Players engage in public and private conversations, form alliances, and vote to eliminate each other' plus topics: benchmark, eval, llm, multi-agent, game, strategy-game.
matched fp:60f59aba066119bf, llm:Repository description and topics: 'A multi-player tournament benchmark that tests LLMs in social reasoning, strategy, and deception. Players engage in public and private conversations, form alliances, and vote to eliminate each other' plus topics: benchmark, eval, llm, multi-agent, game, strategy-game.
matched fp:60f59aba066119bf, llm:Repository description and topics: 'A multi-player tournament benchmark that tests LLMs in social reasoning, strategy, and deception. Players engage in public and private conversations, form alliances, and vote to eliminate each other' plus topics: benchmark, eval, llm, multi-agent, game, strategy-game.