Research map / Image generation / Autoregressive and latent generation
Deep Extrapolation for Attribute-Enhanced Generation
Alvin Chan, Ali Madani, Ben Krause, Nikhil Naik
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 |
|---|---|---|---|
| add cycle-consistency objective (ablation) | GENhance · SST-5 200-Pos | % Strong-Positive: 23.6 -> 49.7 | Improved |
| add cycle-consistency objective (ablation) | GENhance · SST-5 No-Pos | % Strong-Positive: 9.3 -> 21.4 | Improved |
| add latent smoothing objective (ablation) | GENhance w/o smoothing & CC · SST-5 No-Pos | perplexity: 596 -> 126 | Improved |
| add latent smoothing (without CC) (ablation) | GENhance w/o CC · SST-5 200-Pos | % Strong-Positive: 21.5 -> 23.6 | Mixed |
| add latent smoothing objective without CC (ablation) | GENhance w/o Smoothing & CC · ACE2 stability | ddG mean: -7.16 -> -5.59 | Hurt |
| add cycle-consistency on top of smoothing (ablation) | GENhance w/o CC · ACE2 stability | ddG mean: -5.59 -> -7.34 | Improved |
| GENhance vs Gen-Disc baseline | Gen-Disc · SST-5 200-Pos | % Strong-Positive: 26.7 -> 49.7 | Improved |
| GENhance vs Gen-Disc baseline | Gen-Disc · SST-5 200-Pos | perplexity: 63.9 -> 90.5 | Hurt |
| GENhance vs MCMC and CbAS | ACE2 design · ACE2 stability | ddG mean: -4.84 -> -7.34 | Improved |
Closest papers by meaning
- Extrapolative Controlled Sequence Generation via Iterative Refinement (ICML 2023 · similarity 0.89)
- Learning Extrapolative Sequence Transformations from Markov Chains (ICML 2025 · similarity 0.84)
- Diffusion on Language Model Encodings for Protein Sequence Generation (ICML 2025 · similarity 0.82)
- Controllable Protein Sequence Generation with LLM Preference Optimization (AAAI 2025 · similarity 0.82)
- Concept Bottleneck Language Models For Protein Design (ICLR 2025 · similarity 0.81)
- DataGen: Unified Synthetic Dataset Generation via Large Language Models (ICLR 2025 · similarity 0.81)
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.
