Showing papers from Salesforce Research Show all papers
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 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.
Prospective Hindsight uses prediction-reality gaps to weight gradients, improving reinforcement learning performance and self-calibration by targeting blind spots without added objectives.
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.
BehaviorBench evaluates personalized decision modeling using real-world wallet traces across belief and trade prediction tasks, showing personalization improves beliefs more than trades and reveals model failure modes.
Atlanta Poster Session 5, Fri, Dec 11, 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.
Procedural Memory Distillation extracts cross-episode strategy patterns into reusable procedural memory, co-evolving with the policy to improve reasoning benchmarks by up to 13.6% over SDPO.
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.
Variational Policy Distillation co-evolves an adaptive teacher and student via variational EM to extract dense token-level guidance from language feedback, outperforming RLVR and self-distillation baselines on reasoning and code tasks.
Atlanta Poster Session 6, Fri, Dec 11, 4:30 PM–7:30 PM, Hall C1 · Published 2026 · ▲ 12 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.
TerraVis evaluates world-grounded visual consistency in generated images via MLLM workflows, correlating best with human judgments while revealing substantial failures in top models.
Sydney Poster Session 3, Wed, Dec 9, 10:00 AM–1:00 PM, Hall 1-4 · Published 2026 · ▲ 1 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.
Automatic multi-agent systems consistently underperform single-agent chain-of-thought self-consistency despite up to 10x cost, revealing automated architectures suffer from bloat and misaligned complexity rather than true multi-agent benefits.
Sydney Poster Session 2, Tue, Dec 8, 5:00 PM–8:00 PM, Hall 1-4 · Published 2026 · ▲ 1 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.
A Bayes-factor utility framework selects memory turns by evidence of latent preference changes and regulates access accordingly. It outperforms embedding retrieval on preference-intensive long-context dialogue tasks.
Atlanta Poster Session 5, Fri, Dec 11, 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.
OrchRM uses self-supervised orchestration-level reward modeling to train multi-agent orchestrators, cutting token usage by 10x and boosting accuracy up to 8%.
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.