Creates LoCoMo benchmark dataset for very long-term LLM conversational memory and shows current models struggle with lengthy dialogues and long-range temporal dynamics.
hub
When Not to Trust Language Models: Investigating Effectiveness of Parametric and Non-Parametric Memories
35 Pith papers cite this work. Polarity classification is still indexing.
35
Pith papers citing it
hub tools
citation-role summary
dataset 2
citation-polarity summary
roles
dataset 2polarities
use dataset 2representative citing papers
PhantomBench is a new benchmark of 60K+ non-existent terms showing language models hallucinate at rates up to 86.7 percent even when inputs assume the concepts exist.
Activation patching reveals that citation decisions in Llama-3.1-8B RAG are implemented by a distributed attributional ensemble of heads and layers; targeted interventions fix most missed and spurious citations on PopQA.
LatentSkill uses a hypernetwork to generate LoRA adapters from textual skills, enabling weight-space storage that cuts prefill tokens and boosts agent success rates on ALFWorld and Search-QA.
MemTrain introduces two coupled self-supervised proxy tasks on Wikipedia corpora to train general context-memory capabilities in LLMs, reporting gains of up to 17.67 points on long-text and search-based QA benchmarks over direct post-training.
SelSkill applies dual-granularity preference learning to selective skill-or-skip decisions, improving task success by 10.9 points and execution precision by 29.1 points on ALFWorld with Qwen3-8B.
SIOP enables turn-level credit assignment in LLM agents via semantic clustering of final answers as latent outcomes, improving performance on reasoning benchmarks without verifiers.
A normative-descriptive framework shows LLMs' tool-calling perceptions misalign with true need/utility for web search, and hidden-state estimators improve decisions over self-perceived baselines.
SHIFT reformulates neuron editing as learnable gate modulation on under 0.01% parameters to let LLMs adaptively balance contextual and parametric knowledge during RAG generation.
Empirical study of LLM brand recommendations across industries finds moderate concentration (mean Gini 0.28) and low cross-model agreement (41.6%) on top brands.
Misleading tool feedback produces value inversion in LLM agents, with performance dropping below matched no-feedback baselines on HotpotQA and similar tasks.
RISC reformulates self-consistency answer selection as a ranking task solved by a lightweight LambdaRank model with five hand-designed features, yielding better accuracy-efficiency trade-offs than majority voting on QA benchmarks.
RCA is a training-free module that boosts input context signal strength in the residual stream of LLMs by orthogonal decoupling of attention routing from value magnitude.
Sem-ECE is an asymptotically unbiased calibration error estimator for open-ended QA that uses semantic sampling of answers to derive confidence from class frequencies, with two variants that diverge on hard questions.
Decision theory shows that LLM cascades are structurally limited by always incurring the cheap model's cost before deciding to escalate, with the best performance given by the envelope of pairwise cascades rather than fixed chains or many stages.
Verbal-R3 uses a verbal reranker to generate analytic narratives that guide retrieval and reasoning in LLMs, achieving SOTA results on complex QA benchmarks.
R³AG routes queries to retrievers by decomposing capabilities into retrieval quality and generation utility, trained via contrastive learning on document assessments and downstream answer correctness to outperform static methods.
NWCAD uses a two-stream setup with a two-stage gate to prevent accuracy drops on baseline-correct items under non-informative contexts while retaining gains from helpful contexts.
CiPO removes undesired knowledge from both intermediate reasoning steps and final answers in large reasoning models by iteratively optimizing preferences toward valid counterfactual traces while keeping overall reasoning performance intact.
LoVeC uses RL to train LLMs to output verbalized numerical confidence scores for statements in long-form text, achieving better calibration than self-consistency baselines on QA datasets while being 20x faster.
CRAG improves RAG robustness via a retrieval quality evaluator that triggers web augmentation and a decompose-recompose filter to focus on relevant information, yielding better results on short- and long-form generation tasks.
A factorized study finds raw hidden states and attention features hard to beat in-domain for LLM uncertainty probes, but structured compressed features are more robust under distribution shift, with pretrained probes transferring to open-ended generation.
ReCal introduces hierarchical reward decomposition and distribution-aware optimization to address ambiguous credit assignment and optimization bias in RL-based LLM routing.
Early-token log-probabilities from LLM decoding are stronger predictors of reasoning quality than full-sequence statistics in multi-agent debate on essay scoring tasks.
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.
-
PhantomBench: Benchmarking the Non-existential Threat of Language Models
PhantomBench is a new benchmark of 60K+ non-existent terms showing language models hallucinate at rates up to 86.7 percent even when inputs assume the concepts exist.
-
How Do LLMs Cite? A Mechanistic Interpretation of Attribution in Retrieval-Augmented Generation
Activation patching reveals that citation decisions in Llama-3.1-8B RAG are implemented by a distributed attributional ensemble of heads and layers; targeted interventions fix most missed and spurious citations on PopQA.
-
LatentSkill: From In-Context Textual Skills to In-Weight Latent Skills for LLM Agents
LatentSkill uses a hypernetwork to generate LoRA adapters from textual skills, enabling weight-space storage that cuts prefill tokens and boosts agent success rates on ALFWorld and Search-QA.
-
MemTrain: Self-Supervised Context Memory Training
MemTrain introduces two coupled self-supervised proxy tasks on Wikipedia corpora to train general context-memory capabilities in LLMs, reporting gains of up to 17.67 points on long-text and search-based QA benchmarks over direct post-training.
-
Skill or Skip? Learning Selective Skill Invocation in Agentic Tasks via Dual-Granularity Preference Learning
SelSkill applies dual-granularity preference learning to selective skill-or-skip decisions, improving task success by 10.9 points and execution precision by 29.1 points on ALFWorld with Qwen3-8B.
-
Self-Induced Outcome Potential: Turn-Level Credit Assignment for Agents without Verifiers
SIOP enables turn-level credit assignment in LLM agents via semantic clustering of final answers as latent outcomes, improving performance on reasoning benchmarks without verifiers.
-
To Call or Not to Call: A Framework to Assess and Optimize LLM Tool Calling
A normative-descriptive framework shows LLMs' tool-calling perceptions misalign with true need/utility for web search, and hidden-state estimators improve decisions over self-perceived baselines.
-
SHIFT: Gate-Modulated Activation Steering for Knowledge Conflict Mitigation in Retrieval-Augmented Generation
SHIFT reformulates neuron editing as learnable gate modulation on under 0.01% parameters to let LLMs adaptively balance contextual and parametric knowledge during RAG generation.
-
Who Owns the AI Recommendation? A Multi-Industry Empirical Map of Brand Category Ownership Across Large Language Models
Empirical study of LLM brand recommendations across industries finds moderate concentration (mean Gini 0.28) and low cross-model agreement (41.6%) on top brands.
-
Don't Blindly Trust It: How Unreliable Feedback Breaks Tool-Using LLM Agents
Misleading tool feedback produces value inversion in LLM agents, with performance dropping below matched no-feedback baselines on HotpotQA and similar tasks.
-
Boosting Self-Consistency with Ranking
RISC reformulates self-consistency answer selection as a ranking task solved by a lightweight LambdaRank model with five hand-designed features, yielding better accuracy-efficiency trade-offs than majority voting on QA benchmarks.
-
Resonant Context Anchoring: Decoupling Attention Routing and Signal Gain at Inference Time
RCA is a training-free module that boosts input context signal strength in the residual stream of LLMs by orthogonal decoupling of attention routing from value magnitude.
-
A Semantic-Sampling Framework for Evaluating Calibration in Open-Ended Question Answering
Sem-ECE is an asymptotically unbiased calibration error estimator for open-ended QA that uses semantic sampling of answers to derive confidence from class frequencies, with two variants that diverge on hard questions.
-
Is Escalation Worth It? A Decision-Theoretic Characterization of LLM Cascades
Decision theory shows that LLM cascades are structurally limited by always incurring the cheap model's cost before deciding to escalate, with the best performance given by the envelope of pairwise cascades rather than fixed chains or many stages.
-
Verbal-R3: Verbal Reranker as the Missing Bridge between Retrieval and Reasoning
Verbal-R3 uses a verbal reranker to generate analytic narratives that guide retrieval and reasoning in LLMs, achieving SOTA results on complex QA benchmarks.
-
R$^3$AG: Retriever Routing for Retrieval-Augmented Generation
R³AG routes queries to retrievers by decomposing capabilities into retrieval quality and generation utility, trained via contrastive learning on document assessments and downstream answer correctness to outperform static methods.
-
No-Worse Context-Aware Decoding: Preventing Neutral Regression in Context-Conditioned Generation
NWCAD uses a two-stream setup with a two-stage gate to prevent accuracy drops on baseline-correct items under non-informative contexts while retaining gains from helpful contexts.
-
CiPO: Counterfactual Unlearning for Large Reasoning Models through Iterative Preference Optimization
CiPO removes undesired knowledge from both intermediate reasoning steps and final answers in large reasoning models by iteratively optimizing preferences toward valid counterfactual traces while keeping overall reasoning performance intact.
-
LoVeC: Reinforcement Learning for Better Verbalized Confidence in Long-Form Generations
LoVeC uses RL to train LLMs to output verbalized numerical confidence scores for statements in long-form text, achieving better calibration than self-consistency baselines on QA datasets while being 20x faster.
-
Corrective Retrieval Augmented Generation
CRAG improves RAG robustness via a retrieval quality evaluator that triggers web augmentation and a decompose-recompose filter to focus on relevant information, yielding better results on short- and long-form generation tasks.
-
From Signals to Transfer: A Factorised Study of Probe-Based Uncertainty Estimation in Large Language Models
A factorized study finds raw hidden states and attention features hard to beat in-domain for LLM uncertainty probes, but structured compressed features are more robust under distribution shift, with pretrained probes transferring to open-ended generation.
-
ReCal: Reward Calibration for RL-based LLM Routing
ReCal introduces hierarchical reward decomposition and distribution-aware optimization to address ambiguous credit assignment and optimization bias in RL-based LLM routing.
-
Early-Token Confidence Predicts Reasoning Quality in Multi-Agent LLM Debate
Early-token log-probabilities from LLM decoding are stronger predictors of reasoning quality than full-sequence statistics in multi-agent debate on essay scoring tasks.
-
The Confident Liar: Diagnosing Multi-Agent Debate with Log-Probabilities and LLM-as-Judge
In two-agent debate, log-probability confidence aligns with LLM-judged reasoning quality roughly twice as strongly for the Constructor (AUROC 0.804 for critical failure detection) as for the Auditor (0.634).
-
ConMem: Structured Memory-Guided Adaptation in Training-Free Multi-Agent Systems
ConMem distills agent trajectories into structured memory cards organized in a relation-aware graph to enable training-free, relation-coordinated adaptation in LLM-based multi-agent systems.
-
CARE-RL: Capability-Aware Reinforcement Learning for Mitigating Cross-Domain Conflicts
CARE-RL combines PA-GRM for task-adaptive rewards on open-ended tasks and DACSP for modulating RL updates using historical capability directions, reporting higher total average scores than baselines on Qwen models.
-
Only Say What You Know: Calibration-Aware Generation for Long-Form Factuality
Exploration-Commitment Decoupling instantiated as Calibration-Aware Generation improves long-form factuality by up to 13% and reduces decoding time by up to 37% on five benchmarks.
-
Align Documents to Questions: Question-Oriented Document Rewriting for Retrieval-Augmented Generation
QREAM rewrites documents to question-focused style using iterative ICL and distilled FT models, boosting RAG performance by up to 8% relative improvement.
-
IUQ: Interrogative Uncertainty Quantification for Long-Form Large Language Model Generation
IUQ quantifies claim-level uncertainty in long-form LLM generation by combining inter-sample consistency and intra-sample faithfulness through an interrogate-then-respond approach and outperforms baselines on two datasets.
-
A Survey on Hallucination in Large Language Models: Principles, Taxonomy, Challenges, and Open Questions
The paper surveys hallucination in LLMs with an innovative taxonomy, factors, detection methods, benchmarks, mitigation strategies, and open research directions.
-
MATCH: Modulating Attention via In-Context Retrieval for Long-Context Transformers
MATCH augments sparsified attention with an efficient in-context retrieval system to boost performance on long-range recall tasks in transformers.
-
A Reproducible Benchmark and Evidence-Retrieval Software Framework for Silicon Detector R&D Literature
Hybrid sparse-dense retrieval achieves Hit@5 of 0.917 on a new curated benchmark of silicon detector papers with released code and annotations.
-
Retrieve Only Relevant Tables Whether Few or Many: Adaptive Table Retrieval Method
An adaptive thresholding mechanism combined with sliding-window reranking retrieves a query-dependent number of tables from large corpora, improving retrieval and downstream text-to-SQL performance on Spider, BIRD, and Spider 2.0.
- KnowledgeBerg: Evaluating Systematic Knowledge Coverage and Compositional Reasoning in Large Language Models