Research papers, read to you in five minutes.
Research papers, read to you in five minutes.

SMELT tests whether looping an MoE Transformer's middle layers twice actually beats a wider stack, or just quietly spent more FLOPs, parameters, and KV cache. Holding all three budgets equal via narrower width, more experts, and adjusted GQA, it saves 6.8–18.0% training compute, with code and long-context gains compounding at scale.

Agents · Sep 1

StudentSim: Training LLM-based Student Simulators

Simulating individual users for adaptive tutors fails on one of two axes: behavior-cloning captures a learner's mistakes but ignores hints, while prompted frontier LLMs read guidance yet won't reproduce weak-learner errors. StudentSim pools logs to pretrain a LoRA, then specializes per student, hitting 0.51/0.91 fidelity/responsiveness on chess vs GPT-5.4's 0.23/0.72.

Audio/Speech · Aug 31

DreamX-Creator: Democratizing Native Audio-Video Generation at 2K Resolution

Joint audio-video generation usually desyncs on lips and impacts because fusion is applied uniformly across tokens. DreamX-Creator learns a per-token, per-head sigmoid gate on cross-modal attention, and at 7B beats 10-15B open baselines on DeSync by ~40%, though it still trails 22-33B models on raw audio aesthetics.

Diffusion · Aug 29

GenFirst: Generation Before Reconstruction for Stable End-to-End Latent Generative Modeling

Joint VAE+generator training usually collapses, so everyone freezes the VAE and eats a latent space optimized for pixels, not sampling. GenFirst fixes it by splitting KL into prior-fit and entropy terms, then training generation before reconstruction — hitting gFID 0.97 on ImageNet 256 at 2× less compute than REPA-E.

Agents · Aug 28

Agentic Artifact Creation: Systems, Evaluation, Principles, and Opportunities

Agent quality isn't about smarter planners or more self-critique — it's whether the artifact's representation, edit interface, and verifier granularity line up. Across 230 systems, when any link breaks, agents fall back to whole-artifact regeneration. Precise diagnosis is wasted if your only edit is "regenerate."

Never fall behind the literature again.

Free account. Follow topics, build your queue.

Sign up free

Inference Optimization · Sep 1

SMELT: Scaling Laws for Compute-Matched MoE Looped Transformers

SMELT tests whether looping an MoE Transformer's middle layers twice actually beats a wider stack, or just quietly spent more FLOPs, parameters, and KV cache. Holding all three budgets equal via narrower width, more experts, and adjusted GQA, it saves 6.8–18.0% training compute, with code and long-context gains compounding at scale.

Agents · Sep 1

StudentSim: Training LLM-based Student Simulators

Simulating individual users for adaptive tutors fails on one of two axes: behavior-cloning captures a learner's mistakes but ignores hints, while prompted frontier LLMs read guidance yet won't reproduce weak-learner errors. StudentSim pools logs to pretrain a LoRA, then specializes per student, hitting 0.51/0.91 fidelity/responsiveness on chess vs GPT-5.4's 0.23/0.72.

Audio/Speech · Aug 31

DreamX-Creator: Democratizing Native Audio-Video Generation at 2K Resolution

Joint audio-video generation usually desyncs on lips and impacts because fusion is applied uniformly across tokens. DreamX-Creator learns a per-token, per-head sigmoid gate on cross-modal attention, and at 7B beats 10-15B open baselines on DeSync by ~40%, though it still trails 22-33B models on raw audio aesthetics.

Diffusion · Aug 29

GenFirst: Generation Before Reconstruction for Stable End-to-End Latent Generative Modeling

Joint VAE+generator training usually collapses, so everyone freezes the VAE and eats a latent space optimized for pixels, not sampling. GenFirst fixes it by splitting KL into prior-fit and entropy terms, then training generation before reconstruction — hitting gFID 0.97 on ImageNet 256 at 2× less compute than REPA-E.

Agents · Aug 28

Agentic Artifact Creation: Systems, Evaluation, Principles, and Opportunities

Agent quality isn't about smarter planners or more self-critique — it's whether the artifact's representation, edit interface, and verifier granularity line up. Across 230 systems, when any link breaks, agents fall back to whole-artifact regeneration. Precise diagnosis is wasted if your only edit is "regenerate."

Never fall behind the literature again.

Free account. Follow topics, build your queue.

Sign up free

Browse topics

Reasoning 31

Video Generation 27

Inference Optimization 45

Research papers, read to you in five minutes.Research papers, read to you in five minutes.

Research papers, read to you in five minutes.
Research papers, read to you in five minutes.