Sumi is an openly released 7B parameter uniform diffusion language model pretrained from scratch on 1.5T tokens that matches autoregressive models on several benchmarks.
super hub Mixed citations
gpt-oss-120b & gpt-oss-20b Model Card
Mixed citation behavior. Most common role is background (41%).
433
Pith papers citing it
Background
41% of classified citations
abstract
We present gpt-oss-120b and gpt-oss-20b, two open-weight reasoning models that push the frontier of accuracy and inference cost. The models use an efficient mixture-of-expert transformer architecture and are trained using large-scale distillation and reinforcement learning. We optimize the models to have strong agentic capabilities (deep research browsing, python tool use, and support for developer-provided functions), all while using a rendered chat format that enables clear instruction following and role delineation. Both models achieve strong results on benchmarks ranging from mathematics, coding, and safety. We release the model weights, inference implementations, tool environments, and tokenizers under an Apache 2.0 license to enable broad use and further research.
hub tools
citation-role summary
background 33
baseline 16
method 16
other 6
dataset 5
citation-polarity summary
claims ledger
- abstract We present gpt-oss-120b and gpt-oss-20b, two open-weight reasoning models that push the frontier of accuracy and inference cost. The models use an efficient mixture-of-expert transformer architecture and are trained using large-scale distillation and reinforcement learning. We optimize the models to have strong agentic capabilities (deep research browsing, python tool use, and support for developer-provided functions), all while using a rendered chat format that enables clear instruction following and role delineation. Both models achieve strong results on benchmarks ranging from mathematics,
authors
co-cited works
representative citing papers
TW-LegalBench evaluates 13 LLMs on over 30,000 Taiwanese legal tasks from exams and judgments, showing top models pass lawyer thresholds but struggle with exact statute citations.
UltraEP is the first exact-load real-time expert balancer for large-EP MoE training and serving on rack-scale nodes, reaching 94.3% of ideal throughput and 1.49x over no-balancing.
RobotValues is a benchmark of 10K value-conflict scenarios that reveals VLMs default to safety and accommodation while failing to follow instructions to prioritize other values 80% of the time.
Presents the first fully open pipeline for clinical LLMs by unifying eight public QA datasets with three clinician-vetted synthetic extensions and applying it to five base models to achieve benchmark gains while maintaining auditability.
MathAtlas is the first large-scale benchmark for autoformalizing graduate mathematics, where even strong models reach only 9.8% correctness on theorem statements and drop to 2.6% on the hardest dependency-deep subset.
LLMs lack temporal awareness of medical knowledge, showing gradual performance decline on up-to-date facts, much lower accuracy on historical knowledge (25-54% relative), and inconsistent year-to-year predictions.
Sieve dynamically schedules MoE experts across GPU and PIM hardware to handle bimodal token distributions, achieving 1.3x to 1.6x gains in throughput and interactivity over static prior PIM systems on three large models.
Soohak is a 439-problem mathematician-curated benchmark where frontier LLMs reach at most 30.4% on research math challenges and no model exceeds 50% on refusal for ill-posed problems.
MathConstraint generates scalable, automatically verifiable combinatorial problems where LLMs achieve 18.5-66.9% accuracy without tools but roughly double that with solver access.
IRIS-14B is the first LLM trained explicitly for GIMPLE-to-LLVM IR translation and outperforms much larger models by up to 44 percentage points on real-world C code.
RoundPipe achieves near-zero-bubble pipeline parallelism for LLM training on consumer GPUs by dynamically dispatching computation stages round-robin, yielding 1.48-2.16x speedups and enabling 235B model fine-tuning on 8x RTX 4090.
InfiniteScienceGym procedurally generates unbounded scientific repositories with exact ground-truth QA pairs to benchmark LLMs on data reasoning, abstention, and tool use without static datasets.
Large language models display the identifiable victim effect at roughly twice the human baseline, strongly amplified by instruction tuning and chain-of-thought prompting but inverted by reasoning-specialized models.
Tessera performs kernel-granularity disaggregation on heterogeneous GPUs, achieving up to 2.3x throughput and 1.6x cost efficiency gains for large model inference while generalizing beyond prior methods.
The SDE benchmark shows LLMs lag on scientific discovery tasks relative to general science tests, with diminishing scaling returns and shared weaknesses across models.
SpeechCombine produces instruction-following SLMs via speech pre-training followed by direct weight combination with the text LLM instruction delta, without any speech instruction tuning.
OpenSafeIntent benchmark shows models fail to calibrate safety across intent shifts in matched dual-use prompts, indicating current evaluations are insufficient.
A 0.6B LM with length-aware attention adjustments performs competitive in-context retrieval at million-token scale on MS MARCO, NQ, and LIMIT benchmarks.
LLM-generated research ideas cluster more around bridge-like opportunities and synthesis methods than the broader distribution seen in human papers.
ELDR reduces median TPOT by 5.9-13.9% in PD-disaggregated MoE serving via expert signatures from prefill, K-means partitioning, and locality-band routing with KV-co-indexed signature cache.
Introduces GenAI agent framework for auditing personalization algorithms via synthetic accounts with fixed personas, applied to X post-2024 election showing amplification of toxic and right-leaning content varying by ideology.
SABER-Math is an automated benchmark for mathematical IR that uses LLM summaries, topic similarities, and preference tournaments on 283K problems to create reranking tasks, showing embedding models outperform baselines but struggle in symbol-heavy areas and that MTEB does not predict math performanc
LLM agents often fail to abstain at the right time in uncertain multi-turn tasks, and the CONVOLVE context engineering method raises timely abstention rates on WebShop from 26.7 to 57.4 without parameter updates.
citing papers explorer
-
SHIELD: A Diverse Clinical Note Dataset and Distilled Small Language Models for Enterprise-Scale De-identification
SHIELD is a new diverse clinical note dataset paired with distilled small language models that achieve 0.89 span-level precision and 0.88 recall for on-premise PHI de-identification.
-
FlexSQL: Flexible Exploration and Execution Make Better Text-to-SQL Agents
FlexSQL reaches 65.4% on Spider2-Snow by allowing agents to flexibly explore schemas, generate diverse plans, choose SQL or Python execution, and apply two-tiered repair.
-
Spatiotemporal Hidden-State Dynamics as a Signature of Internal Reasoning in Large Language Models
Large reasoning models show measurable hidden-state dynamics that a new statistic can use to distinguish correct reasoning trajectories without labels.
-
MultiBreak: A Scalable and Diverse Multi-turn Jailbreak Benchmark for Evaluating LLM Safety
MultiBreak is a large diverse multi-turn jailbreak benchmark that achieves substantially higher attack success rates on LLMs than prior datasets and reveals topic-specific vulnerabilities in multi-turn settings.
-
When LLMs Stop Following Steps: A Diagnostic Study of Procedural Execution in Language Models
A new benchmark shows LLM first-answer accuracy on procedural arithmetic drops from 63% (5 steps) to 20% (95 steps) due to execution failures like skipped steps and premature answers.
-
Eliminating Hidden Serialization in Multi-Node Megakernel Communication
Perseus removes serialization bottlenecks in multi-node megakernel MoE communication via batched per-destination fences and hardware fence flags, delivering up to 10.3x speedup on proxy transports and matching or exceeding GPU-direct RDMA.
-
Stable-GFlowNet: Toward Diverse and Robust LLM Red-Teaming via Contrastive Trajectory Balance
Stable-GFlowNet stabilizes GFN training for LLM red-teaming by eliminating Z estimation via pairwise comparisons and robust masking against noisy rewards while adding a fluency stabilizer.
-
Why Do LLMs Struggle in Strategic Play? Broken Links Between Observations, Beliefs, and Actions
LLMs encode accurate but brittle internal beliefs about latent game states and convert them poorly into actions, creating systematic gaps that explain strategic failures.
-
Confidence Estimation in Automatic Short Answer Grading with LLMs
A hybrid confidence framework for LLM-based short answer grading combines model signals with aleatoric uncertainty from semantic clustering of responses and improves selective grading reliability over single-source methods.
-
RoadMapper: A Multi-Agent System for Roadmap Generation of Solving Complex Research Problems
RoadMapper multi-agent system improves LLM roadmap generation for complex research problems by over 8% on average and reduces required human expert time by 84%.
-
Test-Time Safety Alignment
Optimizing input embeddings sub-lexically via black-box zeroth-order gradients neutralizes all safety-flagged responses from aligned models on standard benchmarks.
-
VLM-VPI: A Vision-Language Reasoning Framework for Improving Automated Vehicle-Pedestrian Interactions
VLM-VPI uses Qwen3-VL and GPT-OSS models for pedestrian intent and age reasoning plus a tiered safety controller, reporting 92.3% intent accuracy in CARLA and reduced conflicts versus rule-based and supervised baselines.
-
Estimating Tail Risks in Language Model Output Distributions
Importance sampling via activation-steered unsafe proposal models estimates rare harmful-output probabilities in language models with 10-20x fewer samples than brute-force Monte Carlo.
-
Subject-level Inference for Realistic Text Anonymization Evaluation
SPIA is a multi-subject, inference-based benchmark showing that high span-masking rates in text anonymization still leave up to 67% of personal information recoverable through contextual inference.
-
LLaDA2.0-Uni: Unifying Multimodal Understanding and Generation with Diffusion Large Language Model
LLaDA2.0-Uni unifies multimodal understanding and generation inside one discrete diffusion large language model with a semantic tokenizer, MoE backbone, and diffusion decoder.
-
Temporally Extended Mixture-of-Experts Models
Temporally extended MoE layers using the option-critic framework with deliberation costs cut switching rates below 5% while retaining most capability on MATH, MMLU, and MMMLU.
-
TEMPO: Scaling Test-time Training for Large Reasoning Models
TEMPO scales test-time training for large reasoning models by interleaving policy refinement on unlabeled data with critic recalibration on labeled data via an EM formulation, yielding large gains on AIME tasks.
-
Smiling Regulates Emotion During Traumatic Recollection
Smiles during intense negative affect in Holocaust survivor testimonies improve emotional valence trajectories across audio, eye gaze, and text modalities while reducing eye dynamics.
-
Agent-World: Scaling Real-World Environment Synthesis for Evolving General Agent Intelligence
Agent-World autonomously synthesizes verifiable real-world tasks and uses continuous self-evolution to train 8B and 14B agents that outperform proprietary models on 23 benchmarks.
-
SolidCoder: Bridging the Mental-Reality Gap in LLM Code Generation through Concrete Execution
SolidCoder bridges the mental-reality gap in LLM code generation via forced edge-case awareness and concrete sandbox execution, reaching 95.7% pass@1 on HumanEval, 77.0% on CodeContests, and 26.7% on APPS.
-
Agents Explore but Agents Ignore: LLMs Lack Environmental Curiosity
LLM agents discover but largely ignore complete task solutions placed in their environments, revealing a lack of environmental curiosity that persists even with optimized scaffolds.
-
SocialGrid: A Benchmark for Planning and Social Reasoning in Embodied Multi-Agent Systems
SocialGrid benchmark shows even top LLMs achieve below 60% in embodied planning and task completion, with deception detection near random chance regardless of model scale.
-
GroupDPO: Memory efficient Group-wise Direct Preference Optimization
GroupDPO decouples group-wise preference optimization during backpropagation to cut peak memory while keeping the same gradients, allowing larger groups and consistent gains over single-pair DPO plus an NLL term on positives.
-
Prefill-as-a-Service: KVCache of Next-Generation Models Could Go Cross-Datacenter
PrfaaS enables practical cross-datacenter prefill-decode disaggregation for hybrid-attention models via selective offloading, bandwidth-aware scheduling, and cache-aware placement, yielding 54% higher throughput and 64% lower P90 TTFT than homogeneous baselines in a 1T-parameter case study.
-
Development of an LLM-Based System for Automatic Code Generation from HEP Publications
A two-stage LLM system extracts structured analysis selections from HEP papers and references then generates and validates executable code, achieving partial event-level matches on an ATLAS Higgs-to-four-leptons benchmark but limited by hallucination and stochasticity.
-
ELMoE-3D: Leveraging Intrinsic Elasticity of MoE for Hybrid-Bonding-Enabled Self-Speculative Decoding in On-Premises Serving
ELMoE-3D achieves 6.6x average speedup and 4.4x energy efficiency gain for MoE serving on 3D hardware by scaling expert and bit elasticity for elastic self-speculative decoding.
-
Rhetorical Questions in LLM Representations: A Linear Probing Study
Linear probes show rhetorical questions are encoded via multiple dataset-specific directions in LLM representations, with low cross-probe agreement on the same data.
-
Evaluation of Agents under Simulated AI Marketplace Dynamics
Marketplace Evaluation uses repeated-interaction simulations to assess information access systems with marketplace-level metrics such as retention and market share that complement traditional accuracy measures.
-
MetFuse: Figurative Fusion between Metonymy and Metaphor
MetFuse provides the first dataset of 1,000 meaning-aligned quadruplets fusing literal, metonymic, metaphoric, and hybrid sentences, which augments training to boost metonymy and metaphor classification performance on benchmarks.
-
Adaptive Test-Time Scaling for Zero-Shot Respiratory Audio Classification
TRIAGE adaptively scales test-time compute via tiered zero-shot stages for respiratory audio classification, reaching mean AUROC 0.744 across nine tasks while outperforming prior zero-shot methods.
-
A Two-Stage LLM Framework for Accessible and Verified XAI Explanations
A two-stage LLM explainer-verifier framework with iterative refeed improves faithfulness and accessibility of XAI explanations, as shown in experiments across five techniques and three LLM families, with EPR analysis indicating progressive stabilization.
-
Record-Remix-Replay: Hierarchical GPU Kernel Optimization using Evolutionary Search
R^3 optimizes full scientific applications on GPUs better than tuning kernel parameters or compiler flags alone while running nearly an order of magnitude faster than modern evolutionary search methods.
-
Automating Structural Analysis Across Multiple Software Platforms Using Large Language Models
A two-stage multi-agent LLM converts structural inputs to JSON then platform-specific scripts for ETABS, SAP2000, and OpenSees, achieving over 90% accuracy on 20 frame problems across ten trials.
-
SeLaR: Selective Latent Reasoning in Large Language Models
SeLaR selectively applies latent soft reasoning in LLMs via entropy gating and contrastive regularization, outperforming standard CoT on five benchmarks without training.
-
Detecting HIV-Related Stigma in Clinical Narratives Using Large Language Models
An LLM-based NLP tool was developed and tested to identify four types of HIV stigma in clinical notes, achieving up to 0.62 micro F1 score with GatorTron-large.
-
A Multi-Stage Validation Framework for Trustworthy Large-scale Clinical Information Extraction using Large Language Models
A multi-stage framework with prompt calibration, rule-based filtering, semantic checks, judge LLM review, and predictive validation enables trustworthy LLM extraction of substance use disorder diagnoses from nearly 920,000 clinical notes, achieving F1 of 0.80 and superior care-engagement prediction.
-
Do Domain-specific Experts exist in MoE-based LLMs?
Domain-specific experts exist in MoE-based LLMs, and the training-free DSMoE framework steers them to outperform baselines on target domains with no added inference cost.
-
Watch Before You Answer: Learning from Visually Grounded Post-Training
Filtering post-training data to visually grounded questions improves VLM video understanding performance by up to 6.2 points using 69% of the data.
-
$\pi^2$: Structure-Originated Reasoning Data Improves Long-Context Reasoning Ability of Large Language Models
A data curation pipeline turning Wikipedia tables into verified multi-hop reasoning examples improves long-context reasoning performance in LLMs through supervised fine-tuning.
-
Formalized Information Needs Improve Large-Language-Model Relevance Judgments
Synthetically formalizing information needs into topics with descriptions and narratives improves LLM relevance assessor agreement with humans and reduces over-labeling of relevant documents on TREC Deep Learning and Robust04.
-
Learning Robust Visual Features in Computed Tomography Enables Efficient Transfer Learning for Clinical Tasks
VoxelFM learns robust 3D CT visual features via DINO self-distillation that transfer effectively to seven clinical task categories using frozen backbones and lightweight heads, outperforming prior CT foundation models even on report generation.
-
Structured Multi-Criteria Evaluation of Large Language Models with Fuzzy Analytic Hierarchy Process and DualJudge
Fuzzy AHP and DualJudge deliver more stable and calibrated LLM evaluations than direct scoring by breaking assessments into explicit criteria and adaptively fusing intuitive and deliberative judgments.
-
InCoder-32B-Thinking: Industrial Code World Model for Thinking
InCoder-32B-Thinking uses error-feedback synthesized thinking traces and a code world model to reach top open-source scores on general and industrial code benchmarks including 81.3% on LiveCodeBench and 84.0% on CAD-Coder.
-
On the Role of Reasoning Patterns in the Generalization Discrepancy of Long Chain-of-Thought Supervised Fine-Tuning
SFT on divergent branch-heavy CoT from DeepSeek-R1 yields worse generalization than convergent CoT from gpt-oss despite lower loss, but filtering frequent branches improves average performance by 3.6% on five reasoning benchmarks.
-
LLM-Generated Fault Scenarios for Evaluating Perception-Driven Lane Following in Autonomous Edge Systems
A decoupled offline-online framework uses LLMs and latent diffusion models to generate fault scenarios for testing edge-based lane-following models, revealing large robustness drops under conditions like fog.
-
Learning to Retrieve from Agent Trajectories
Retrievers trained on agent trajectories via the LRAT framework improve evidence recall, task success, and efficiency in agentic search benchmarks.
-
Rethinking Language Model Scaling under Transferable Hypersphere Optimization
HyperP transfers optimal learning rates across model width, depth, tokens, and MoE granularity under Frobenius-sphere constraints, delivering stable scaling and 1.58x efficiency gains.
-
Ensemble-Based Uncertainty Estimation for Code Correctness Estimation
Ensemble Semantic Entropy improves correlation with code correctness over single-model methods and powers a cascading scaling system that cuts FLOPs by 64.9% while preserving performance on LiveCodeBench.
-
A Domain-Specific Language for LLM-Driven Trigger Generation in Multimodal Data Collection
A DSL combined with LLMs generates consistent, low-latency triggers for selective multimodal sensor data collection, outperforming direct code generation in consistency and speed with comparable detection performance.
-
Tracking Capabilities for Safer Agents
AI agents can generate code in a capability-safe Scala dialect that statically prevents information leakage and malicious side effects while preserving task performance.