Creates LoCoMo benchmark dataset for very long-term LLM conversational memory and shows current models struggle with lengthy dialogues and long-range temporal dynamics.
super hub Mixed citations
Scalable training of
Mixed citation behavior. Most common role is unclear (64%).
160
Pith papers citing it
unclear
64% of classified citations
hub tools
citation-role summary
background 9
other 2
citation-polarity summary
claims ledger
- background vey of graph meets large language model: progress and future directions. InProceedings of the Thirty- Third International Joint Conference on Artificial Intelligence, pages 8123-8131. Andrés Montoyo, Patricio Martínez-Barco, and Alexan- dra Balahur. 2012. Subjectivity and sentiment analy- sis: An overview of the current state of the area and envisaged developments.Decision Support Systems, 53(4):675-679. Bo Pang, Lillian Lee, and Shivakumar Vaithyanathan. 2002. Thumbs up? sentiment classificatio
- background whether u is semantically broader than k. The two samples are expressed as X={x i}n i=1 andY={y j}m j=1,(3) where xi =x u,i and yj =x k,j, with n, m fixed (typically, we subsample to a common size to con- trol the variance across words). A natural null hypothesis is that the two words have the same dispersion but different mean directions. H0 :disp(X) =disp(Y) withE[X]̸=E[Y]allowed.(4) This is because the mean direction is a strong nui- sance factor in contextual embedding spaces. Even if two wo
- background domain. Given Xv ∈R Sv×Dv and Xt ∈R St×Dt, the goal is to refine Xv by aggregating contextual information across scales. We define N scales with two adapter sets: G= {G1, . . . ,GN } (MGFA) and C={C 1, . . . ,CN } (MCFA). At each scale n, features are reshaped to a grid X (0) v ∈R H×W×D v and downsampled by Down(·,2 n−1): X (n) v = Down(X(0) v ,2 n−1).(4) Let Xv,n = Seq(X (n) v ) denote the flattened se- quence. We then refine and fuse: Gn =G n(Xv,n), C n =C n(Xv,n, Xt),(5) ˜Xv,n =G n +w C n,(6)
- other Question: Eukaryotic genes tend to consist of coding regions (exons) and non-coding regions (introns). The figure shows how such a gene leads to the production of a protein. Which of the following statements is true? A. Thymine content of (1) and (2) is approximately equal. B. The process occurring between (2) and (3) takes place in the cytosol. C. (4) can hybridise with (2). D. The number of amino acid residues in (5) must equal the number of nucleotide residues in (2). E. All processes occurri
- background Question: Eukaryotic genes tend to consist of coding regions (exons) and non-coding regions (introns). The figure shows how such a gene leads to the production of a protein. Which of the following statements is true? A. Thymine content of (1) and (2) is approximately equal. B. The process occurring between (2) and (3) takes place in the cytosol. C. (4) can hybridise with (2). D. The number of amino acid residues in (5) must equal the number of nucleotide residues in (2). E. All processes occurri
- other sharing & image reaction functions are integrated to add a multi-modal dimension to the long-term dialogues.2 The image sharing function is called when the agent decides to send an image. This process includes: (1) Generate a caption c for the intended image using M; (2) Convert the caption c into relevant keywords w using M; (3) Use the keywords k to find an image through web search W EB(k)3; (4) Share the chosen image. Con- versely, the image reaction function is triggered upon receiving an im
authors
co-cited works
representative citing papers
Invisible Unicode perturbations, optimized from surrogate compressors then adapted by prior-guided evolution under a low query budget, cause large information loss in agent context compression without changing human-visible text.
P2R decouples perception from reasoning in VLMs via a two-stage process and PRA-GRPO alternating RL training, reporting gains such as 93.2% on V-Star for the 4B model over its Qwen3-VL backbone.
MECoBench is a benchmark showing that multimodal agent collaboration improves embodied task performance when communication balances coordination costs, with gains also under noisy conditions.
The paper proposes an operator-level visual-token skipping framework for MLLMs that reduces TFLOPs by 33.7% on Qwen3-VL while retaining 99.5% performance across VQA benchmarks.
SMDA fits ridge regression on SAE features to distill symbolic policies then decomposes each SFT example's influence via feature-activation and output-probability deltas, demonstrated on refusal behavior in Llama-3.2-3B-Instruct.
SEATauBench is the first agent benchmark for SEA languages, finding that performance holds for language-only changes but degrades sharply with full domain localization.
Agentic CLEAR automates multi-level evaluation of LLM agents, generating textual insights at system, trace, and node granularity that align with human annotations and predict task success.
Introduces Causal Functional Signatures grounded in causal evidence and ILP-learned architectural signatures to enable explicit, comparable, and portable mechanistic claims across model scales.
An 8B autoregressive LM implements a language-switching backdoor via a three-phase circuit with early trigger composition, orthogonal mid-layer propagation, and final-layer MLP conversion, routed through a single-position serial bottleneck.
R3-Streaming uses cascaded control with age-aware memory forgetting and TB-GRPO reinforcement learning to reach SOTA scores of 57.92 on OVO-Bench and 76.36 on StreamingBench with 95-96% fewer visual tokens.
New metrics KSS and KPS are introduced to evaluate multilingual machine unlearning quality and cross-language consistency in LLMs, addressing limitations of single-language evaluation protocols.
LongBEL improves biomedical entity linking consistency by combining full-document context with memory of previous predictions trained via cross-validation rather than gold labels.
A new benchmark dataset drawn from Japan's National Assessment of Academic Ability supplies real exam layouts, diagrams, Japanese text, and nationwide student response distributions for evaluating multimodal LLMs.
Semantic Softmax aggregates probabilities from semantic synonyms around target labels to correct renormalization bias in zero-shot LLM classification, lowering calibration error and raising AUROC and F1.
CA-SQL achieves 51.72% execution accuracy on the challenging tier of the BIRD benchmark using GPT-4o-mini by scaling exploration breadth according to estimated task difficulty, evolutionary prompt seeding, and candidate voting.
A new permutation test uses Householder reflection to align word embedding clouds before testing dispersion differences, cutting Type-I error by 32.5% and speeding up 23x on GPU.
Introduces the GeoDial dataset of 1.3K multimodal geometry tutoring dialogs grounded in diagram highlights, proposes an annotation protocol, and shows that fine-tuned VLMs improve dialog but struggle with accurate highlights.
LOVER creates an unsupervised logic-regularized verifier that reaches 95% of supervised verifier performance on reasoning tasks across 10 datasets.
POSTCONDBENCH is a new multilingual benchmark that evaluates LLM postcondition generation on real code using defect discrimination to assess completeness beyond surface matching.
S²R² improves robustness of LoRA-tuned LLMs to prompt perturbations by penalizing semantic-segment drift while preserving clean performance and cross-dataset transfer.
Proposes self-function vectors and a controlled evaluation protocol to quantify aleatoric uncertainty in ICL separately from epistemic uncertainty for more reliable LLM confidence measures.
OptiVerse is a new benchmark spanning neglected optimization domains that shows LLMs suffer sharp accuracy drops on hard problems due to modeling and logic errors, with a Dual-View Auditor Agent proposed to improve performance.
SpanDec achieves competitive NER accuracy with improved efficiency by using a final-stage lightweight decoder for span representations and early candidate filtering to reduce redundant computation.
citing papers explorer
-
Evaluating Very Long-Term Conversational Memory of LLM Agents
Creates LoCoMo benchmark dataset for very long-term LLM conversational memory and shows current models struggle with lengthy dialogues and long-range temporal dynamics.
-
Out of Sight: Compression-Aware Content Protection against Agentic Crawlers
Invisible Unicode perturbations, optimized from surrogate compressors then adapted by prior-guided evolution under a low query budget, cause large information loss in agent context compression without changing human-visible text.
-
Perceive-to-Reason: Decoupling Perception and Reasoning for Fine-Grained Visual Reasoning
P2R decouples perception from reasoning in VLMs via a two-stage process and PRA-GRPO alternating RL training, reporting gains such as 93.2% on V-Star for the 4B model over its Qwen3-VL backbone.
-
MECoBench: A Systematic Study of Multimodal Agent Collaboration in Embodied Environments
MECoBench is a benchmark showing that multimodal agent collaboration improves embodied task performance when communication balances coordination costs, with gains also under noisy conditions.
-
Attend, Transform, or Silence: Operator-Level Visual Skipping for Efficient Multimodal LLM Inference
The paper proposes an operator-level visual-token skipping framework for MLLMs that reduces TFLOPs by 33.7% on Qwen3-VL while retaining 99.5% performance across VQA benchmarks.
-
Symbolic Mechanistic Data Attribution: Tracing Training Influence to Learned Behavioral Policies
SMDA fits ridge regression on SAE features to distill symbolic policies then decomposes each SFT example's influence via feature-activation and output-probability deltas, demonstrated on refusal behavior in Llama-3.2-3B-Instruct.
-
SEATauBench: Adapting Tool-Agent-User Evaluation Into Low-Resource Southeast Asian Languages
SEATauBench is the first agent benchmark for SEA languages, finding that performance holds for language-only changes but degrades sharply with full domain localization.
-
Agentic CLEAR: Automating Multi-Level Evaluation of LLM Agents
Agentic CLEAR automates multi-level evaluation of LLM agents, generating textual insights at system, trace, and node granularity that align with human annotations and predict task success.
-
From Circuit Evidence to Mechanistic Theory: An Inductive Logic Approach
Introduces Causal Functional Signatures grounded in causal evidence and ILP-learned architectural signatures to enable explicit, comparable, and portable mechanistic claims across model scales.
-
Language-Switching Triggers Take a Latent Detour Through Language Models
An 8B autoregressive LM implements a language-switching backdoor via a three-phase circuit with early trigger composition, orthogonal mid-layer propagation, and final-layer MLP conversion, routed through a single-position serial bottleneck.
-
An Efficient Streaming Video Understanding Framework with Agentic Control
R3-Streaming uses cascaded control with age-aware memory forgetting and TB-GRPO reinforcement learning to reach SOTA scores of 57.92 on OVO-Bench and 76.36 on StreamingBench with 95-96% fewer visual tokens.
-
Knowledge Beyond Language: Bridging the Gap in Multilingual Machine Unlearning Evaluation
New metrics KSS and KPS are introduced to evaluate multilingual machine unlearning quality and cross-language consistency in LLMs, addressing limitations of single-language evaluation protocols.
-
LongBEL: Long-Context and Document-Consistent Biomedical Entity Linking
LongBEL improves biomedical entity linking consistency by combining full-document context with memory of previous predictions trained via cross-validation rather than gold labels.
-
Human-Grounded Multimodal Benchmark with 900K-Scale Aggregated Student Response Distributions from Japan's National Assessment of Academic Ability
A new benchmark dataset drawn from Japan's National Assessment of Academic Ability supplies real exam layouts, diagrams, Japanese text, and nationwide student response distributions for evaluating multimodal LLMs.
-
The Silent Vote: Improving Zero-Shot LLM Reliability by Aggregating Semantic Neighborhoods
Semantic Softmax aggregates probabilities from semantic synonyms around target labels to correct renormalization bias in zero-shot LLM classification, lowering calibration error and raising AUROC and F1.
-
CA-SQL: Complexity-Aware Inference Time Reasoning for Text-to-SQL via Exploration and Compute Budget Allocation
CA-SQL achieves 51.72% execution accuracy on the challenging tier of the BIRD benchmark using GPT-4o-mini by scaling exploration breadth according to estimated task difficulty, evolutionary prompt seeding, and candidate voting.
-
Accurate and Efficient Statistical Testing for Word Semantic Breadth
A new permutation test uses Householder reflection to align word embedding clouds before testing dispersion differences, cutting Type-I error by 32.5% and speeding up 23x on GPU.
-
GeoDial: A Multimodal Conversational Tutoring Dataset for Geometry Problem-Solving with Visual Tutor Turns
Introduces the GeoDial dataset of 1.3K multimodal geometry tutoring dialogs grounded in diagram highlights, proposes an annotation protocol, and shows that fine-tuned VLMs improve dialog but struggle with accurate highlights.
-
Logic-Regularized Verifier Elicits Reasoning from LLMs
LOVER creates an unsupervised logic-regularized verifier that reaches 95% of supervised verifier performance on reasoning tasks across 10 datasets.
-
POSTCONDBENCH: Benchmarking Correctness and Completeness in Formal Postcondition Inference
POSTCONDBENCH is a new multilingual benchmark that evaluates LLM postcondition generation on real code using defect discrimination to assess completeness beyond surface matching.
-
Where Do Prompt Perturbations Break Generation? A Segment-Level View of Robustness in LoRA-Tuned Language Models
S²R² improves robustness of LoRA-tuned LLMs to prompt perturbations by penalizing semantic-segment drift while preserving clean performance and cross-dataset transfer.
-
Quantifying Aleatoric Uncertainty of In-Context Learning for Robust Measure of LLM Prediction Confidence
Proposes self-function vectors and a controlled evaluation protocol to quantify aleatoric uncertainty in ICL separately from epistemic uncertainty for more reliable LLM confidence measures.
-
OptiVerse: A Comprehensive Benchmark towards Optimization Problem Solving
OptiVerse is a new benchmark spanning neglected optimization domains that shows LLMs suffer sharp accuracy drops on hard problems due to modeling and logic errors, with a Dual-View Auditor Agent proposed to improve performance.
-
Decoding Text Spans for Efficient and Accurate Named-Entity Recognition
SpanDec achieves competitive NER accuracy with improved efficiency by using a final-stage lightweight decoder for span representations and early candidate filtering to reduce redundant computation.
-
ATIR: Towards Audio-Text Interleaved Contextual Retrieval
Defines ATIR task and benchmark for mixed audio-text queries; MLLM model with token compression shows substantial gains over strong baselines.
-
Memory-Augmented LLM-based Multi-Agent System for Automated Feature Generation on Tabular Data
MALMAS is a memory-augmented multi-agent LLM system that generates diverse, high-quality features for tabular data via agent decomposition, routing, and iterative memory-guided refinement.
-
Text-to-Distribution Prediction with Quantile Tokens and Neighbor Context
Quantile tokens inserted into LLM inputs combined with neighbor retrieval enable direct prediction of full distributions, yielding lower MAPE and narrower intervals than baselines on Airbnb and StackSample tasks.
-
Indic-CodecFake meets SATYAM: Towards Detecting Neural Audio Codec Synthesized Speech Deepfakes in Indic Languages
Introduces the Indic-CodecFake dataset for Indic codec deepfakes and SATYAM, a novel hyperbolic ALM that outperforms baselines through dual-stage semantic-prosodic fusion using Bhattacharya distance.
-
Structure Guided Retrieval-Augmented Generation for Factual Queries
SG-RAG frames retrieval as subgraph matching to ensure LLMs meet every condition in factual queries and reports large gains over baselines on a new 120k-pair ERQA dataset.
-
From Experience to Skill: Multi-Agent Generative Engine Optimization via Reusable Strategy Learning
MAGEO is a multi-agent system that distills validated editing patterns into reusable optimization skills for generative engines, outperforming heuristic baselines on visibility and fidelity via a new benchmark and evaluation protocol.
-
Cell-Based Representation of Relational Binding in Language Models
Large language models encode relational bindings via a cell-based representation: a low-dimensional linear subspace in which each cell corresponds to an entity-relation index pair and attributes are retrieved from the matching cell.
-
LQM: Linguistically Motivated Multidimensional Quality Metrics for Machine Translation
LQM introduces a six-level linguistically motivated error taxonomy for MT evaluation and applies it via expert annotation to LLM outputs on a new 3,850-sentence multi-dialect Arabic corpus.
-
Region-Grounded Report Generation for 3D Medical Imaging: A Fine-Grained Dataset and Graph-Enhanced Framework
Introduces the first large-scale 3D PET/CT dataset with fine-grained RoI annotations for Vietnamese and a graph-enhanced HiRRA framework that achieves SOTA report generation by modeling RoI dependencies.
-
Prosody as Supervision: Bridging the Non-Verbal--Verbal for Multilingual Speech Emotion Recognition
NOVA-ARC is a hyperbolic geometry framework that transfers emotion supervision from labeled non-verbal vocalizations to unlabeled verbal speech in multiple languages via optimal transport prototype alignment and consistency regularization.
-
Self-Consistency from Only Two Samples: CoT-PoT Ensembling for Efficient LLM Reasoning
Cross-modal agreement between chain-of-thought and program-of-thought reasoning enables self-consistency with only two LLM samples, reducing sampling cost by 9.3x while improving accuracy.
-
GaLa: Hypergraph-Guided Visual Language Models for Procedural Planning
GaLa uses hypergraph representations of objects and a TriView encoder with contrastive learning to improve vision-language models on procedural planning benchmarks.
-
HeadRank: Decoding-Free Passage Reranking via Preference-Aligned Attention Heads
HeadRank lifts preference optimization into attention space via entropy-regularized head selection and distribution regularizers to sharpen discriminability for efficient listwise reranking.
-
Stress-Testing the Reasoning Competence of LLMs With Proofs Under Minimal Formalism
ProofGrid is a new benchmark for LLM reasoning that uses machine-checkable proofs in minimal formal notation, revealing progress on basic tasks but major gaps in complex combinatorial and synthesis reasoning.
-
CODI: Compressing Chain-of-Thought into Continuous Space via Self-Distillation
CODI compresses explicit CoT into continuous space via self-distillation and is the first implicit method to match explicit CoT performance on GSM8k at GPT-2 scale with 3.1x compression and 28.2% higher accuracy than prior implicit approaches.
-
Longformer: The Long-Document Transformer
Longformer uses local windowed attention plus task-specific global attention to achieve linear scaling and state-of-the-art results on long-document language modeling, QA, and summarization after pretraining.
-
Token-Flow Firewall: Semantic Runtime Auditing for Persistent AI Agents
TokenWall mediates persistent-agent security by auditing source–sink token flows with a local small model and selective large-model escalation, cutting CIK-Bench attack success to 12.5% at low benign latency.
-
Tool-Making and Self-Evolving LLM Agents in Low-Latency Systems
Compiling repeated SOP nodes into environment-grounded, versioned tools cuts production p50 latency by 42% and end-to-end error rate by up to 53% in a 44-node fulfillment-center alarm-triage agent.
-
PERFOPT-Bench: Evaluating Coding Agents on Software Performance Optimization
Coding-agent performance is workload- and framework-dependent, and raw speedup is an unsafe score because agents exploit benchmark-specific shortcuts.
-
Measuring the practice of shared-decision making (OPTION12): An Investigation into Open-sourced Smaller LLMs (OS-sLLMs) for Better Privacy and Sustainability
General-domain OS-sLLMs moderately correlate with human OPTION12 SDM scores on Dutch melanoma transcripts; medical models fail via hallucination, and a Judge-LLM consensus is proposed.
-
PluraMath: Extending Mathematical Reasoning Evaluation Beyond High-Resource Languages
PluraMath extends PolyMath with human-validated math problems in 18 mid-to-extreme low-resource languages and benchmarks 27 reasoning LLMs, finding a persistent high- vs low-resource performance gap.
-
MemDefrag: Latent Memory Defragmentation for Large Language Models
Ranking latent-memory fragments by middle-layer attention density and keeping only the top few more than doubles long-term knowledge-retention accuracy (43.0% vs. 17.4%/17.6%) versus MemoryLLM and M+.
-
On the Limits of Steering Vectors for Preference-Aligned Generation
Empirical evaluation on the PLUME benchmark shows steering vectors vary widely in trait expressibility, degrade on task transfer, and lose effectiveness when multiple vectors are composed.
-
Persona Non Grata: LLM Persona-Driven Generations in MCQA are Unstable in Distinct Dimensions
Persona-driven generations by LLMs in MCQA tasks exhibit instability that differs systematically by model family, size, domain, and prompt format.
-
MetaHOPE: A Metaphor-Oriented Evaluation Framework for Analysing MT and LLM Translation Errors
MetaHOPE frames metaphor translation quality through five severity-weighted error types and, in a pilot, finds metaphor-related errors dominate MT/LLM translation penalties.
-
YOMI-Bench: A Benchmark for Evaluating Kanji Reading and Phonological Understanding of LLMs for Japanese
YOMI-Bench is a new benchmark of four tasks for kanji reading and phonological understanding in LLMs, showing low performance even for Japanese-specific and commercial models.