Peer ballot
Rank ten papers.
Then rank them again.
Put the same ten papers in order twice, once by expected research impact and once by the talks you would attend. Nothing you enter leaves this browser.
A practice set of ten historical papers
The ten are generative-model papers from ICLR 2017, described by plain-language working titles. Bylines are hidden. The papers carry no TenPeers acceptance, and none implies that its authors took part.
Your orders stay in this browser, so a reload picks up where you left off.
- Ballot 1Which papers will other researchers build on and cite? Order the ten by expected research impact over the next two years, most influential first.
- Ballot 2Which talks would you go to in person? The same ten papers in a fresh shuffle. Put the one you would most want to hear first.
The ten papers
Ranking here needs JavaScript. Without it, this is the list of papers with their source records.
Training a generator against a future discriminator
To update the generator, the method simulates several steps that the discriminator would take next. The experiments test whether this makes GAN training more stable.
Source recordA different likelihood for image pixels
An autoregressive image model represents pixel values with a discretized mixture of logistic distributions. The authors examine how that change and architectural revisions affect training and image likelihood.
Source recordLearning a density by reversing a transformation
Invertible coupling layers transform a simple distribution into a more complex one. The model supports sampling and exact likelihood evaluation in the same architecture.
Source recordDifferentiating through a categorical choice
A soft version of a categorical sample allows gradients to pass through a discrete choice during training. A temperature parameter controls how closely the sample resembles a one-hot decision.
Source recordA continuous distribution for discrete variables
The authors define a family of relaxed discrete random variables with a tractable density. They use it to estimate gradients and test the method on density estimation and structured prediction.
Source recordLearning an encoder and a generator together
A discriminator compares data–latent pairs from an encoder to pairs from a generator. Both networks learn through the same adversarial objective.
Source recordGiving the discriminator an energy function
The discriminator learns to assign lower energy to regions near the data. The generator searches for examples that the discriminator also assigns low energy.
Source recordMeasuring the likelihood of a decoder’s samples
The authors use annealed importance sampling to evaluate decoder-based generative models. They study the gap between convincing samples and likelihood estimates.
Source recordWhy adversarial training can stall
This paper analyzes instability and saturation in GAN training. Targeted experiments investigate when the discriminator leaves the generator with an uninformative gradient.
Source recordSeparating factors of variation in images
A variational autoencoder changes the weight on its latent constraint. The authors study how that trade-off affects the recovery of separate visual factors.
Source record
What this shows, and what it doesn't
It shows the interaction: two separate orders of the same ten papers, and how far they can differ. It does not test whether your impact order predicts later citations. These are historical papers, some familiar to experienced researchers, so a prospective test on unfamiliar papers would say more.
No cohort exists yet and this page collects no votes. For how orders are combined and tested, read the method.