Showing papers from ETH Zürich Show all papers
Sydney Poster Session 1, Tue, Dec 8, 10:00 AM–1:00 PM, Hall 1-4 · Published 2026
Readers and the AI panel: vote on this paper to see what they said.
Only vote on papers you've read. Sign in with GitHub to vote.
Sydney Poster Session 4, Wed, Dec 9, 5:00 PM–8:00 PM, Hall 1-4 · Published 2026
Readers and the AI panel: vote on this paper to see what they said.
Only vote on papers you've read. Sign in with GitHub to vote.
Paris Poster Session 2, Wed, Dec 9, 5:00 PM–7:00 PM, Paris Poster Hall · Published 2026
Readers and the AI panel: vote on this paper to see what they said.
Only vote on papers you've read. Sign in with GitHub to vote.
Sydney Poster Session 3, Wed, Dec 9, 10:00 AM–1:00 PM, Hall 1-4 · Published 2026
Readers and the AI panel: vote on this paper to see what they said.
Only vote on papers you've read. Sign in with GitHub to vote.
Sydney Poster Session 6, Thu, Dec 10, 5:00 PM–8:00 PM, Hall 1-4 · Published 2026
Readers and the AI panel: vote on this paper to see what they said.
Only vote on papers you've read. Sign in with GitHub to vote.
Sydney Poster Session 2, Tue, Dec 8, 5:00 PM–8:00 PM, Hall 1-4 · Published 2026
Readers and the AI panel: vote on this paper to see what they said.
Only vote on papers you've read. Sign in with GitHub to vote.
Language generation in the limit is recast as recall-precision trade-offs, showing that allowing infinitely many vanishing-frequency hallucinations can strictly increase recall when adversaries withhold target portions.
Sydney Poster Session 6, Thu, Dec 10, 5:00 PM–8:00 PM, Hall 1-4 · Published 2026
Readers and the AI panel: vote on this paper to see what they said.
Only vote on papers you've read. Sign in with GitHub to vote.
PGLD optimizes synthesis-aware stochastic DNA libraries via policy gradients to bypass synthesis cost limits, enabling million-sequence libraries for antibody exploration at low cost.
Sydney Poster Session 5, Thu, Dec 10, 10:00 AM–1:00 PM, Hall 1-4 · Published 2026
Readers and the AI panel: vote on this paper to see what they said.
Only vote on papers you've read. Sign in with GitHub to vote.
Targeted Full Conformal Prediction uses vision-language models to prune labels and scale full conformal image classification with stable coverage and modest overhead.
Paris Poster Session 1, Wed, Dec 9, 12:30 PM–2:30 PM, Paris Poster Hall · Published 2026
Readers and the AI panel: vote on this paper to see what they said.
Only vote on papers you've read. Sign in with GitHub to vote.
PaGeR adapts perspective 3D foundation models to panoramas to predict depth, normals, and sky masks in one pass, achieving state-of-the-art 360-degree geometry estimation.
Sydney Poster Session 4, Wed, Dec 9, 5:00 PM–8:00 PM, Hall 1-4 · Published 2026 · ▲ 4 on Hugging Face · Code ★ 157
Readers and the AI panel: vote on this paper to see what they said.
Only vote on papers you've read. Sign in with GitHub to vote.
LittleLearner is a 5B-parameter model trained on grade-capped elementary data to study controlled knowledge acquisition and bounded capability growth.
Sydney Poster Session 5, Thu, Dec 10, 10:00 AM–1:00 PM, Hall 1-4 · Published 2026 · ▲ 9 on Hugging Face
Readers and the AI panel: vote on this paper to see what they said.
Only vote on papers you've read. Sign in with GitHub to vote.
Under local PŁ conditions, unique optimistic lower-level selection ensures hyper-gradient differentiability via pseudoinverses, yielding HG-MS with manifold-dependent convergence and strong LLM reweighting results.
Paris Poster Session 3, Thu, Dec 10, 12:30 PM–2:30 PM, Paris Poster Hall · Published 2026
Readers and the AI panel: vote on this paper to see what they said.
Only vote on papers you've read. Sign in with GitHub to vote.
A byte-level sequential Monte Carlo algorithm samples from composed language model ensembles, outperforming naive probability averaging across structured generation tasks.
Sydney Poster Session 1, Tue, Dec 8, 10:00 AM–1:00 PM, Hall 1-4 · Published 2026
Readers and the AI panel: vote on this paper to see what they said.
Only vote on papers you've read. Sign in with GitHub to vote.
Class Adaptive Conformal Training adaptively shapes class-conditional prediction sets via augmented Lagrangian optimization without distributional assumptions, yielding smaller sets with valid coverage.
Atlanta Poster Session 3, Thu, Dec 10, 10:00 AM–1:00 PM, Hall C1 · Published 2026
Readers and the AI panel: vote on this paper to see what they said.
Only vote on papers you've read. Sign in with GitHub to vote.
StitchVM stitches pretrained pixel-space reward models onto frozen diffusion backbones to build accurate noisy-latent value functions for efficient diffusion alignment, accelerating DPS 3.2× and DiffusionNFT 2.3×.
Sydney Poster Session 3, Wed, Dec 9, 10:00 AM–1:00 PM, Hall 1-4 · Published 2026 · ▲ 10 on Hugging Face
Readers and the AI panel: vote on this paper to see what they said.
Only vote on papers you've read. Sign in with GitHub to vote.
EchoPrune treats redundant video tokens as temporal echoes and prunes them via query relevance and reconstruction error, letting VideoLLMs process up to 20x more frames for +8.6% accuracy and 5.6x faster prefilling.
Sydney Poster Session 5, Thu, Dec 10, 10:00 AM–1:00 PM, Hall 1-4 · Published 2026
Readers and the AI panel: vote on this paper to see what they said.
Only vote on papers you've read. Sign in with GitHub to vote.
Anchor PCA finds shared low-rank directions across domains by trading variance for cross-domain agreement, yielding robust embeddings that generalize to unseen domains.
Paris Poster Session 1, Wed, Dec 9, 12:30 PM–2:30 PM, Paris Poster Hall · Published 2026
Readers and the AI panel: vote on this paper to see what they said.
Only vote on papers you've read. Sign in with GitHub to vote.
Lumberjack improves differentially private random forests via heavy hitter pruning of deep trees, achieving state-of-the-art privacy-utility trade-offs.
Sydney Poster Session 4, Wed, Dec 9, 5:00 PM–8:00 PM, Hall 1-4 · Published 2026
Readers and the AI panel: vote on this paper to see what they said.
Only vote on papers you've read. Sign in with GitHub to vote.
A discrete-time stochastic control formulation yields value-driven transport policies that generate data via straight, fast, robust paths and support conditional generation and guidance.
Sydney Poster Session 6, Thu, Dec 10, 5:00 PM–8:00 PM, Hall 1-4 · Published 2026
Readers and the AI panel: vote on this paper to see what they said.
Only vote on papers you've read. Sign in with GitHub to vote.
Discrete diffusion models learn data support before frequencies because reverse edits scale by validity first and coefficients second; absorbing diffusion prioritizes validity-improving moves over uniform diffusion's trichotomy.
Sydney Poster Session 5, Thu, Dec 10, 10:00 AM–1:00 PM, Hall 1-4 · Published 2026
Readers and the AI panel: vote on this paper to see what they said.
Only vote on papers you've read. Sign in with GitHub to vote.