All papers

This paper demonstrates that frozen short-context drafting models can effectively serve ultra-long context speculative decoding without fine-tuning on long sequences.

The paper presents a benchmark and harness to delegate routine tool choices from large language models to cheaper decision models.

The paper measures the information gap between coding agents' internal execution logs and their user-facing progress reports.

The paper presents a benchmark evaluating whether conversational agents agree on verifiable plans that satisfy both parties' private conditions.

Proposes NEAR, a training-free framework using neighborhood-derived residual representations to reduce retrieval errors from shared non-target information in memory records.

The paper identifies a linear spatial workspace subspace in multimodal language models that encodes continuous 3D coordinate representations.

Proposes a method linking execution traces to specific skill segments to guide automated revision of software engineering agent skills.

Gates reinforcement learning policy-shift distillation using intermediate checkpoints to ensure student models only learn consistent teacher shifts.

The paper introduces a benchmark to measure how tool-calling agents perform when interacting with emotionally varied, impatient users.

The paper evaluates harnesses for legislative document conversion, showing structured multi-agent pipelines narrow performance gaps between models.

The paper introduces a tool-augmented reinforcement learning framework for evidence-grounded chest X-ray diagnostic reasoning.

The paper analyzes geometric properties of text embeddings to predict how many nearest neighbors an individual example loses under dimensionality reduction.

This paper proposes a rubric-based reward model that derives criteria strictly from instructions and evaluates responses pointwise using tool assistance.

The paper analyzes the runtime performance differences of test-passing issue resolution patches generated by language models on SWE-bench.

Proposes explicit learning of physical state transitions between current and future scenes to train robust vision-language-action robotic policies.

This paper evaluates how reliably typed decision models switch to fallback options and remain consistent under equivalent decision rules.

The paper presents a benchmark evaluating whether conversational agents agree on verifiable plans that satisfy both parties' private conditions.

Uses teacher critiques on incorrect student rollouts to improve on-policy distillation and recover supervision during mathematical reasoning tasks.

This paper uses limited human preference signals to calibrate automated evaluation metrics for biomedical retrieval-augmented question answering.

This paper introduces a metric to estimate the compatibility of pretrained models before running costly alignment or checkpoint merging.

Presents a multi-dimensional data pruning framework that filters code repositories by syntax, diversity, and representation for efficient retriever fine-tuning.

The paper proposes a training-free framework that uses uncertainty signals to escalate computer-use tasks from weak models to strong models.

This paper introduces CommerceGen-Studio, an end-to-end framework using specialized agents and diffusion models to generate e-commerce product images.

The paper proposes an execution harness that validates intermediate results before allowing scientific agent workflows to release outputs.

The paper presents an intervention method using Shapley values to assign credit to individual actions within multi-action agentic retrieval sets.

The paper proposes ReMaRk, a reinforcement learning method that trains recommender models using all available reader replies simultaneously to reduce variance.

This paper introduces UnCoBench to benchmark how language model agents handle unhelpful, incomplete, or missing communications from partner agents.

Applies regenerative software primitives so programs can be deleted and rebuilt by agents without losing architectural intent and knowledge.

Allocates computation unevenly across passes in looped mixture-of-experts models to improve efficiency without losing performance.

The paper proposes an agentic framework enabling multimodal models to identify information gaps and retrieve in-context demonstrations iteratively.

The paper introduces an executable evaluation platform and benchmark to measure contextual integrity and privacy compliance in multi-step language model agents.

The paper presents a code auditing method that uses language models to infer developer beliefs and flag contradictions as potential vulnerabilities.

This paper evaluates language-model-generated disaster situation reports against human-written reports using a domain-specific sixteen-slot information schema.

This paper introduces UnCoBench to benchmark how language model agents handle unhelpful, incomplete, or missing communications from partner agents.

The paper proposes VEDJE, compressing video features into reusable caches with feature-change prediction for efficient and scalable video-text retrieval.

Introduces a set-level reinforcement learning objective to train language models to produce diverse and document-grounded scientific research ideas.

The paper reveals vulnerabilities in prompt-predictability metrics used to estimate human contribution in AI-assisted writing.

This paper introduces a reinforcement learning method that calibrates test pass-rate rewards to create adaptive training curricula for software engineering agents.

This paper presents a foundation model and a hundred-million-image dataset for mask-based object, text, and watermark erasure in images.

This paper audits whether decision models and generative language models can serve as valid synthetic respondents for psychometric surveys.

The paper proposes a vision-language-action model where action execution avoids inference latency by decoupling from explicit reasoning output at runtime.

Introduces an asynchronous prefix denoising framework for extreme latent compression in diffusion models.

Uses a multi-codebook diffusion language model to jointly represent text and discrete visual tokens for multimodal generation.

The paper proposes OpCredit, a policy learning framework that assigns individual operation credits for agentic knowledge graph question answering.

This paper examines the verifiability gap in quantized model supply chains, showing that common GGUF artifacts lack necessary build environment disclosures.

The paper presents SPROUT, a snapshot-based rollout framework to scale reinforcement learning for long-horizon software engineering tasks.

The authors identify a derivation gap where LLM judges reject correct tool arguments lacking direct textual references, and propose a dedicated reference model.

The authors identify a derivation gap where LLM judges reject correct tool arguments lacking direct textual references, and propose a dedicated reference model.

Proposes a synthetic self-distillation method using a slightly larger aligner to enable continual post-training in small language models.

The paper surveys memory-centric compression techniques for managing inference resources and context windows in large language models.

The paper presents a benchmark that isolates agent memory evolution from downstream retrieval and generation failures.

This paper proposes an agentic video framework combining change-aware memory and targeted replay for streaming video understanding.

The paper evaluates embedding distance metrics to detect generalization boundaries and distribution shifts in AI-based code vulnerability models.

This paper proposes verbal regularization prompts for on-policy self-distillation to prevent teacher models from taking shortcut reasoning paths.

The paper audits an enterprise confidence scoring mechanism for text-to-SQL models on the BIRD and Spider benchmarks.

Presents a framework alternating multi-agent sessions across context resets to overcome reasoning degradation in long-horizon machine learning engineering.

Presents a multi-stage search system that retrieves and ranks relevant scientific literature directly from informal research ideas.

The paper demonstrates that a single adversarial agent can cause benign models in multi-agent discussions to quietly adopt false beliefs.

Evaluates whether failure feedback helps self-improving agents fix specific target bugs rather than general errors.

This paper introduces a recursive self-improvement method that unifies instruction-following and instruction generation within a shared model for vision-language navigation.