Discover the Best AI Resources

The First Fable 5.1 Reasoning Data

Contains 5,000 coding and chain-of-thought reasoning traces generated by Fable 5.1 — ~150M tokens of step-by-step programming CoT. Deduplicated and filtered for high quality; intended for supervised fine-tuning and distillation to improve reasoning in smaller models.

MoreThought/Fable-5.1-Max-Reasoning-Filtered-1000x

A curated set of 1,000 high-quality chain-of-thought coding and reasoning traces generated by Fable 5.1, totaling ~30M tokens (109 MB). Designed for SFT/distillation to teach smaller models step-by-step programmatic reasoning and debugging.

VDN-Minimax-H3

Adds a plug-and-play linear-attention branch and LoRA adapters to MiniMax-H3 to run text-to-video generation faster than real-time (near-lossless quality tradeoffs). Includes an optimized FP8 inference stack and a community license with regional restrictions.

TikTok Videos, 4.5 Billion

Provides 4.5 billion TikTok video records with captions, timestamps, music IDs and engagement counts for research; split across 27 zstd-compressed Parquet files (~289 GB) and sampled via TikTok's mobile API; released for research-use only with privacy and ToS caveats.

SolarWM: Open Data and Scalable Training for Long-Horizon Video World Models

Provides an end-to-end, reproducible foundation for camera‑controllable, long‑horizon video world models — converting 1.43M clips from 10 datasets into a unified canonical corpus and releasing data, pipelines, recipes, and weights. Introduces backbone‑native adaptation and a three‑stage training recipe to produce 5B–33B models that enable minute‑to‑hour rollouts after training on 5s sequences.

Language Models Can Control Their Own Attention

Introduces Declarative Attention (DA), a zero-shot protocol that has LMs declare which parts of long context to attend to during chain-of-thought, letting the runtime build dynamic attention masks and skip most KV-cache reads. Produces large token savings (up to ~52% on Gemma-4-31B) with modest accuracy loss.

Iris: Climbing to the Search Frontier

Presents two LLM-based search agents (Iris-mini and Iris-pro) trained by alternating supervised fine-tuning and reinforcement learning against live web search. Key features: web-graph-derived multi-hop tasks with entity abstraction, SFT–RL climbing, inference-time context management, and state-of-the-art open-source benchmark results.

Terminal-Universe: Turning Agent Trajectories into Scalable Terminal Environments

Reconstructs executable terminal workspaces from recorded agent trajectories and synthesizes verifiable single- and multi-round coding tasks for agent training; it replays file operations, uses an LLM completion agent to fill missing files/dependencies, and verifies tasks with autogenerated test suites.

Unlocking Lossless Speedups in LLMs via Discrete Diffusion

Uses lightweight discrete-diffusion adapters to draft blocks of tokens in parallel while keeping the original autoregressive model as the final arbiter, enabling lossless speedups (up to 3×) and compatibility with existing AR weights via a Diffusion Distillation phase.

Rethinking On-Policy Distillation of Large Language Models II: One Training Example

Studies on-policy distillation (OPD) at the data-minimal limit by training on a single query, measuring state coverage and alignment dynamics, and showing OPD is often data-overfed but algorithm-starved.

Compile by Training: Turning Natural-Language Specifications into Local Neural Functions

Converts a natural-language function specification into a reusable local neural function by using teacher models to synthesize examples and finetuning a small adapter for a compact interpreter. Achieves higher semantic accuracy (83.6% on FuzzyBench-Hard) at the cost of roughly one minute compile time; produces versionable PAW artifacts for local deployment.

FlowBalance: Verifier-Grounded Self-Improvement from On-Policy Reasoning Experience

Combines sparse verifier outcomes with dense privileged‑hindsight token scoring to learn an outcome‑calibrated, normalized distribution over complete responses for on‑policy self‑improvement. Key features: sign‑gated guidance (retain/reverse/disable per verifier advantage), profiled trajectory balance with one log‑partition per rollout group, and explicit correction against false‑positive self‑guidance.