Studies when and how to combine visual future rollouts from world models with abstract reasoning in multimodal LLMs. Proposes PF-OPSD — a teacher-student distillation that uses ground-truth future videos during training — and evaluates on two human-verified benchmarks, improving accuracy ≈10% while improving robustness to noisy rollouts.
AI Dataset2026
Generates synthetic coding-agent session traces by pairing remotely hosted open agent models with local llama.cpp user models across real open-source codebases. Each trace records read/write/edit/bash actions and tool use; the dataset is a reproducible cartesian product (20×3×20×20 = 24,000 sessions) under an MIT license.
AI Agent Papers2026
Represents episodic memory as a Cue–Tag–Content graph and integrates LLM reasoning into active retrieval so agents iteratively reconstruct and prune evidence paths for long-horizon questions. Reports up to 23% gains on LoCoMo / LongMemEval while reducing token and runtime costs.
Generates repository-specific LoRA adapters via a hypernetwork to inject repo-level knowledge into code LMs with zero inference-time token overhead. Provides a Static snapshot mode and an Evo mode that updates adapters per commit; evaluated on the 604-repo RepoPeftBench.
AI Agent Papers2026
Measures how coding agents explore repositories by asking them to return a ranked, line-level list of code regions relevant to an issue under a fixed line budget. Covers 848 issues across 203 repos and 10 languages; evaluates coverage, ranking, and context-efficiency to isolate exploration quality.
Removes the subspace of frequent, uninformative tokens that LLMs inject into text embeddings via the model's unembedding matrix. EmbedFilter is a lightweight linear transform that refines LLM-derived embeddings to improve zero‑shot semantic retrieval, enable dimensionality reduction, and speed up indexing; code on GitHub.
AI Model2026
Code-focused sparse Mixture-of-Experts LLM designed for agentic coding and terminal/tool use, offering very long context (256K) and long outputs. Released with open weights under Apache-2.0 and optimized for transformers/vLLM workflows.
AI Agent Papers2026
Benchmark for long-horizon computer-use agents that must orchestrate GUI, CLI, and code operations within single trajectories across 114 real-world tasks. Evaluated on a real Ubuntu desktop and paired with a trajectory-aware judge that inspects deliverables, artifacts, and action traces—revealing a top PassRate of ~41.2%.
Standardizes representation-level evaluation for tabular encoders by exporting row-, column-, and table-level embeddings and probing them with shared lightweight heads across three suites (TRL-CTbench, TRL-Rbench, TRL-DLTE). Supplies curated benchmark assets and task rewrites (50 OpenML tables, 123 targets, a 47,772-table DLTE lake) to enable fair cross-paradigm comparison.
Generates outcome-specific, dialectical rationales with an LLM and derives continuous, calibrated risk scores for irregularly sampled medical time series—mitigating risk polarization. Reports +3.3% average AUPRC and 81% reduction in calibration error across three benchmarks; code released.
AI Model2026
Implements MXFP4 quantization on MoE experts plus a BF16 DFlash block-diffusion drafter to propose whole-token blocks for verification, cutting memory bandwidth and backbone forward passes for trillion‑parameter text generation—targeting long‑context, agent and code workloads.
AI Agent Papers2026
Turns raw datasets into verifiable multimodal news features via a multi-agent newsroom pipeline. Key innovations: (1) an Inspector that links each claim to data/code/external references for re-execution and audit; (2) multimodal asset generation (interactive maps, audio, visuals) tailored to the story.