Research map / Reinforcement learning / Visual and representation RL
Pre-Trained Language Models for Interactive Decision-Making
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 changed | Where | Measured | Effect |
|---|---|---|---|
| Use natural-language text inputs instead of arbitrary sequential encodings (ablation) | LID · VirtualHome Novel Tasks, 500 demos | success rate: 18.0 $\pm$ 3.8 -> 22.2 $\pm$ 1.7 | Helped |
| Use natural-language text inputs instead of arbitrary sequential encodings (ablation) | LID · VirtualHome Novel Tasks, 20K demos | success rate: 57.8 $\pm$ 0.9 -> 58.2 $\pm$ 1.6; 58.8 $\pm$ 0.9 | No clear effect |
| Use sequential inputs rather than pooled fixed-dimensional vectors (ablation) | LID · VirtualHome Novel Tasks | success rate: 2.0 $\pm$ 0.6 -> 58.2 $\pm$ 2.3 | Helped |
| Initialize policy weights with a pretrained LM rather than training from scratch (ablation) | LID · VirtualHome Novel Tasks | success rate: 47.0 $\pm$ 2.8 -> 58.2 $\pm$ 2.3 | Helped |
| Fine-tune pretrained LM weights rather than freezing them (ablation) | LID · VirtualHome Novel Tasks | success rate: 17.0 $\pm$ 2.9 -> 58.2 $\pm$ 2.3 | Helped |
| Use text encoding rather than convolutional observation encoding (ablation) | LID · BabyAI GoToRedBall, 100 demos | success rate: 92.5 -> 93.9 | Helped |
| Use text encoding rather than convolutional observation encoding (ablation) | LID · BabyAI GoToLocal, 100 demos | success rate: 69.5 -> 64.6 | Hurt |
| Use text encoding rather than convolutional observation encoding (ablation) | LID · BabyAI PickupLoc, 100 demos | success rate: 25.0 -> 28.7 | Helped |
| Use text encoding rather than convolutional observation encoding (ablation) | LID · BabyAI PutNextLocal, 100 demos | success rate: 17.9 -> 11.1 | Hurt |
| Initialize an online PPO policy from LID-ADG | PPO · VirtualHome In-Distribution | success rate: +53.7 % | Helped |
Closest papers by meaning
- Enhancing Decision-Making of Large Language Models via Actor-Critic (ICML 2025 · similarity 0.90)
- On the Modeling Capabilities of Large Language Models for Sequential Decision Making (ICLR 2025 · similarity 0.89)
- EvoLM: In Search of Lost Language Model Training Dynamics (NeurIPS 2025 · similarity 0.89)
- Can large language models explore in-context? (NeurIPS 2024 · similarity 0.89)
- Representation-Based Exploration for Language Models: From Test-Time to Post-Training (ICLR 2026 · similarity 0.88)
- Large Language Models are Interpretable Learners (ICLR 2025 · similarity 0.88)
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.