AI Paper Atlas

Research map / Multi-agent systems and games / Human-AI and goal-conditioned

Incorporating Pragmatic Reasoning Communication into Emergent Language

Yipeng Kang, Tonghan Wang, Gerard de Melo

NeurIPS 2020 · Multi-agent systems and games › Human-AI and goal-conditioned · approach: Emergent communication in referential games · 9 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
Add two-sided IBR pragmatic reasoningEmergent language agents · Referential game, Challengeaccuracy: 53.0$\pm$2.6 -> 80.6$\pm$2.8Improved
Use GameTable-sequential equilibrium selectionGameTable pragmatic model · Referential game, Challengeaccuracy: 74.6$\pm$2.0 -> 94.0$\pm$0.6Improved
Use probabilistic RSA instead of deterministic IBRTwo-sided pragmatic reasoning · Referential game, Challengeaccuracy: 80.6$\pm$2.8 -> 55.9$\pm$1.9Hurt
Use two-round IBR instead of converged IBR (ablation)IBR pragmatic model · Referential game, Overallaccuracy: 95.5$\pm$0.6 -> 95.6$\pm$0.5No clear effect
Use two-round RSA instead of converged RSA (ablation)RSA pragmatic model · Referential game, Overallaccuracy: 91.9$\pm$0.9 -> 93.1$\pm$0.8Hurt
Give the speaker no weight for consistency with its prior (SampleL λ=0) (ablation)SampleL · Referential game, Overallaccuracy: 90.1$\pm$0.9 -> 90.1$\pm$1.1No clear effect
Use one-sided SampleL rather than baseline decodingEmergent language agents · Referential game, Challengeaccuracy: 53.0$\pm$2.6 -> 54.8$\pm$2.9No clear effect
Use GameTable instead of IBRTwo-sided pragmatic reasoning · Referential game, Challengeaccuracy: 80.6$\pm$2.8 -> 74.6$\pm$2.0Hurt
Use a reconstructed virtual opponent model rather than an exact copyIBR · Referential game with virtual opponentaccuracy: 80.6$\pm$2.8 -> 68.6$\pm$1.6Hurt
Use a reconstructed virtual opponent model rather than an exact copyGameTable · Referential game with virtual opponentaccuracy: 74.6$\pm$2.0 -> 58.1$\pm$2.2Hurt
Use a reconstructed virtual opponent model rather than an exact copyGameTable-sequential · Referential game with virtual opponentaccuracy: 94.0$\pm$0.6 -> 69.9$\pm$1.8Hurt
Add pragmatic reasoning to communicationStarCraft II ally agents · 1o2r_vs_4rbattle win rate under message droppingImproved

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy