Research map / Large language models / Language, safety and interpretability
Agentic Collaboration as an Information Bottleneck Problem
Shizhe He, Avanika Narayan, Ishan S. Khare, Scott Linderman, Christopher Re, Dan Biderman
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 changed | Where | Measured | Effect |
|---|---|---|---|
| scale compressor from 1.5B to 7B | Qwen-2.5 compressor · LongHealth | accuracy: 1.6× more accurate | Improved |
| scale compressor from 1-1.5B to 7-8B | compressor-predictor system · FinanceBench | accuracy: 2.6× more accurate | Improved |
| scale compressor size from 1.5B to 7B | Qwen-2.5 compressor · LongHealth | output length: 4.6× more concise | Improved |
| scale compressor size from 1.5B to 7B | Qwen-2.5 compressor · LongHealth | FLOPs-per-generation: 1.3% more FLOPs | No clear effect |
| scale predictor from 70B to 405B | Llama-3 predictor · LongHealth | accuracy: 12% improvement | Mixed |
| scale predictor from 70B to 405B | Llama-3 predictor · FinanceBench | accuracy: 1% improvement | No clear effect |
| instruct compressor to output 3, 6 or 9 sentences (ablation) | Qwen-2.5 compressors · LongHealth | accuracy and mutual information | No clear effect |
| use Qwen-2.5 compressor family instead of Llama | compressor · LongHealth and FinanceBench | correctness (logistic regression) | Improved |
| pair predictor with compressor of the same family | compressor-predictor system · LongHealth | distortion | No clear effect |
| add larger Qwen-2.5-14B compressor with GPT-4o predictor instead of uncompressed context | GPT-4o predictor · DeepResearch Bench | RACE score: 2.3% higher RACE at 28.1% of cost | Improved |
| raise mutual information rate of compressor output | compressor-predictor system · FineWeb | perplexity: r = -0.84, R2 = 0.71 | Improved |
Closest papers by meaning
- Can Compressed LLMs Truly Act? An Empirical Evaluation of Agentic Capabilities in LLM Compression (ICML 2025 · similarity 0.91)
- Fix the Structural Bottleneck: Context Compression via Explicit Information Transmission (NeurIPS 2026 · similarity 0.89)
- When Less is More: The LLM Scaling Paradox in Context Compression (NeurIPS 2026 · similarity 0.88)
- Latent Collaboration in Multi-Agent Systems (ICML 2026 · similarity 0.88)
- AMA-Bench: Evaluating Long-Horizon Memory for Agentic Applications (ICML 2026 · similarity 0.88)
- When Parallelism Pays Off: Cohesion-Aware Task Partitioning for Multi-Agent Coding (NeurIPS 2026 · 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.
