Hashimoto , title =

representative citing papers

citing papers explorer

Showing 45 of 45 citing papers.

• WildChat: 1M ChatGPT Interaction Logs in the Wild cs.CL · 2024-05-02 · accept · none · ref 32

  WildChat releases a dataset of 1 million ChatGPT conversations with timestamps, demographics, and headers, claimed to be the most diverse and multilingual such resource available.

• ORPO: Monolithic Preference Optimization without Reference Model cs.CL · 2024-03-12 · conditional · none · ref 102

  ORPO performs preference alignment during supervised fine-tuning via a monolithic odds ratio penalty, allowing 7B models to outperform larger state-of-the-art models on alignment benchmarks.

• PoisonForge: Task-Level Targeted Poisoning Benchmark for Instruction-Tuned LLMs cs.CR · 2026-05-22 · accept · none · ref 3

  PoisonForge benchmark shows that 1% poisoned examples achieve over 70% attack success rate on targeted tasks across 11 of 12 tested LLMs with under 0.5% leakage to non-target tasks.

• Where Pretraining writes and Alignment reads: the asymmetry of Transformer weight space cs.LG · 2026-05-15 · unverdicted · none · ref 39

  Pretraining and alignment induce asymmetric geometric traces in transformer weights because alignment updates concentrate in read pathways due to activation covariance while write pathways inherit less structure from alignment losses.

• Enjoy Your Layer Normalization with the Computational Efficiency of RMSNorm cs.LG · 2026-05-14 · conditional · none · ref 101

  A framework to identify and convert foldable layer normalizations to RMSNorm for exact equivalence and faster inference in deep neural networks.

• AnchorSeg: Language Grounded Query Banks for Reasoning Segmentation cs.CV · 2026-04-20 · unverdicted · none · ref 56

  AnchorSeg uses ordered query banks of latent reasoning tokens plus a spatial anchor token and a Token-Mask Cycle Consistency loss to achieve 67.7% gIoU and 68.1% cIoU on the ReasonSeg benchmark.

• CODI: Compressing Chain-of-Thought into Continuous Space via Self-Distillation cs.CL · 2025-02-28 · unverdicted · none · ref 37

  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.

• SWE-RL: Advancing LLM Reasoning via Reinforcement Learning on Open Software Evolution cs.SE · 2025-02-25 · unverdicted · none · ref 80

  SWE-RL uses RL on software evolution data to train LLMs achieving 41% on SWE-bench Verified with generalization to other reasoning tasks.

• Refusal in Language Models Is Mediated by a Single Direction cs.LG · 2024-06-17 · accept · none · ref 38

  Refusal in language models is mediated by a single direction in residual stream activations that can be erased to disable safety or added to elicit refusal.

• Magpie: Alignment Data Synthesis from Scratch by Prompting Aligned LLMs with Nothing cs.CL · 2024-06-12 · unverdicted · none · ref 35

  Magpie synthesizes 300K high-quality alignment instructions from Llama-3-Instruct via auto-regressive prompting on partial templates, enabling fine-tuned models to match official instruct performance on AlpacaEval, ArenaHard, and WildBench.

• Self-Rewarding Language Models cs.CL · 2024-01-18 · conditional · none · ref 23

  Iterative self-rewarding via LLM-as-Judge in DPO training on Llama 2 70B improves instruction following and self-evaluation, outperforming GPT-4 on AlpacaEval 2.0.

• Q-Align: Teaching LMMs for Visual Scoring via Discrete Text-Defined Levels cs.CV · 2023-12-28 · conditional · none · ref 69

  Q-Align trains LMMs on discrete text-defined levels for visual scoring, achieving SOTA on IQA, IAA, and VQA while unifying the tasks in OneAlign.

• EvoPrompt: Connecting LLMs with Evolutionary Algorithms Yields Powerful Prompt Optimizers cs.CL · 2023-09-15 · unverdicted · none · ref 127

  EvoPrompt uses LLMs to run evolutionary operators on populations of prompts, outperforming human-engineered prompts by up to 25% on BIG-Bench Hard tasks across 31 datasets.

• LIMA: Less Is More for Alignment cs.CL · 2023-05-18 · conditional · none · ref 12

  Fine-tuning a 65B model on 1,000 high-quality examples produces output that humans rate as good as or better than GPT-4 in 43% of cases, indicating most capabilities come from pretraining.

• LLaMA-Adapter: Efficient Fine-tuning of Language Models with Zero-init Attention cs.CV · 2023-03-28 · conditional · none · ref 174

  LLaMA-Adapter turns frozen LLaMA 7B into a capable instruction follower using only 1.2M new parameters and zero-init attention, matching Alpaca while extending to image-conditioned reasoning on ScienceQA and COCO.

• kNNGuard: Turning LLM Hidden Activations into a Training-Free Configurable Guardrail cs.LG · 2026-07-02 · conditional · none · ref 21

  Multi-layer Fisher-weighted kNN over frozen-LLM activations, fused with embedding kNN, yields competitive F1 guardrails from a 50-example bank with no fine-tuning and sub-10-second domain adaptation.

• Distill to Detect: Exposing Stealth Biases in LLMs through Cartridge Distillation cs.CL · 2026-07-01 · unverdicted · none · ref 37

  D2D distills distributional shifts between a suspected model and its base into a cartridge adapter to amplify and detect stealth biases in LLMs across multiple types.

• The Heterogeneous Safety Impacts of Benign Multilingual Fine-Tuning cs.CL · 2026-06-27 · unverdicted · none · ref 78

  Benign multilingual fine-tuning causes language-specific safety drifts with adversarial compliance rates rising up to four-fold, decoupled from capability gains.

• Can LLMs Reliably Self-Report Adversarial Prefills, and How? cs.CL · 2026-06-22 · conditional · none · ref 41

  LLMs rarely recognize their own prefilled (adversarially forced) outputs, and the recognition they do show tracks refusal circuits and question phrasing rather than a robust sense of self.

• HMARS: A Hierarchical Multi-Agent Memory System for Long-Context Reasoning cs.IR · 2026-06-03 · unverdicted · none · ref 41

  HMARS introduces a hierarchical multi-agent memory system that outperforms standard retrieval and other baselines on long-document and multi-turn reasoning tasks through improved evidence coverage.

• Assisted Counterspeech Writing at the Crossroads of Hate Speech and Misinformation cs.CL · 2026-05-21 · conditional · none · ref 157

  LLMs generate adequate counterspeech for co-occurring hate and misinformation in 40% of cases, with a mixed knowledge strategy from fact-checkers and NGOs proving most effective after expert revision.

• A Single Neuron Is Sufficient to Bypass Safety Alignment in Large Language Models cs.CL · 2026-05-08 · unverdicted · none · ref 28

  Suppressing one refusal neuron or amplifying one concept neuron bypasses safety alignment in LLMs from 1.7B to 70B parameters without training or prompt engineering.

• Minimizing Collateral Damage in Activation Steering cs.LG · 2026-05-01 · unverdicted · none · ref 23

  Activation steering is cast as constrained optimization that minimizes collateral damage by weighting perturbations according to the empirical second-moment matrix of activations instead of assuming isotropy.

• InvEvolve: Evolving White-Box Inventory Policies via Large Language Models with Performance Guarantees cs.LG · 2026-05-01 · unverdicted · none · ref 98 · 2 links

  InvEvolve evolves inventory policies using LLMs with RL and provides statistical safety guarantees, outperforming classical and DL methods on synthetic and real data.

• Different Paths to Harmful Compliance: Behavioral Side Effects and Mechanistic Divergence Across LLM Jailbreaks cs.CR · 2026-04-20 · unverdicted · none · ref 18

  Different LLM jailbreak techniques achieve similar harmful compliance but lead to distinct behavioral side effects and mechanistic changes.

• Flex Attention: A Programming Model for Generating Optimized Attention Kernels cs.LG · 2024-12-07 · unverdicted · none · ref 17

  FlexAttention supplies a compiler-driven interface that expresses common attention variants in a few lines of PyTorch and emits optimized kernels whose speed matches hand-written implementations.

• Inference Scaling Laws: An Empirical Analysis of Compute-Optimal Inference for Problem-Solving with Language Models cs.AI · 2024-08-01 · conditional · none · ref 6

  Empirical analysis shows scaling inference compute via strategies like tree search can be more efficient than scaling model parameters, with 7B models plus novel search outperforming 34B models.

• EAGLE: Speculative Sampling Requires Rethinking Feature Uncertainty cs.LG · 2024-01-26 · unverdicted · none · ref 34

  EAGLE resolves feature-level uncertainty in speculative sampling via one-step token advancement, delivering 2.7x-3.5x speedup on LLaMA2-Chat 70B and doubled throughput across multiple model families and tasks.

• The Falcon Series of Open Language Models cs.CL · 2023-11-28 · conditional · none · ref 211

  Falcon-180B is a 180B-parameter open decoder-only model trained on 3.5 trillion tokens that approaches PaLM-2-Large performance at lower cost and is released with dataset extracts.

• Chain-of-Verification Reduces Hallucination in Large Language Models cs.CL · 2023-09-20 · unverdicted · none · ref 73

  Chain-of-Verification reduces hallucinations in large language models by drafting responses, planning independent verification questions, answering them separately, and generating a final verified output.

• DoLa: Decoding by Contrasting Layers Improves Factuality in Large Language Models cs.CL · 2023-09-07 · conditional · none · ref 73

  DoLa reduces hallucinations in LLMs by contrasting logits from later versus earlier layers during decoding, improving truthfulness on TruthfulQA by 12-17 absolute points without fine-tuning or retrieval.

• Enhancing Chat Language Models by Scaling High-quality Instructional Conversations cs.CL · 2023-05-23 · conditional · none · ref 200

  UltraChat supplies 1.5 million high-quality multi-turn dialogues that, when used to fine-tune LLaMA, produce UltraLLaMA, which outperforms prior open-source chat models including Vicuna.

• GiVA: Gradient-Informed Bases for Vector-Based Adaptation cs.CL · 2026-04-23 · unverdicted · none · ref 20

  GiVA uses gradients to initialize vector adapters so they match LoRA performance at eight times lower rank while keeping extreme parameter efficiency.

• FedProxy: Federated Fine-Tuning of LLMs via Proxy SLMs and Heterogeneity-Aware Fusion cs.LG · 2026-04-21 · unverdicted · none · ref 60

  FedProxy replaces weak adapters with a proxy SLM for federated LLM fine-tuning, outperforming prior methods and approaching centralized performance via compression, heterogeneity-aware aggregation, and training-free fusion.

• mPLUG-Owl3: Towards Long Image-Sequence Understanding in Multi-Modal Large Language Models cs.CV · 2024-08-09 · unverdicted · none · ref 64

  mPLUG-Owl3 introduces hyper attention blocks to integrate vision and language for long image-sequence understanding and reports SOTA results on single-image, multi-image, and video benchmarks.

• Aligning Modalities in Vision Large Language Models via Preference Fine-tuning cs.LG · 2024-02-18 · unverdicted · none · ref 101

  POVID generates AI-created preference data to fine-tune vision-language models with DPO, reducing hallucinations and improving benchmark scores.

• GradShield: Alignment Preserving Finetuning cs.CL · 2026-05-13 · unverdicted · none · ref 44

  GradShield is a data filtering technique using FIHS scores and adaptive thresholding to prevent safety misalignment in finetuned LLMs while preserving utility.

• Qwen-Scope: Turning Sparse Features into Development Tools for Large Language Models cs.CL · 2026-05-12 · unverdicted · none · ref 4

  Qwen-Scope provides open-source sparse autoencoders for Qwen models that function as practical interfaces for steering, evaluating, data workflows, and optimizing large language models.

• Learning in the Fisher Subspace: A Guided Initialization for LoRA Fine-Tuning cs.LG · 2026-05-01 · unverdicted · none · ref 44

  Fisher information from the target data distribution supplies a task-dependent criterion for selecting LoRA directions that outperforms weight-magnitude heuristics.

• Step-Video-T2V Technical Report: The Practice, Challenges, and Future of Video Foundation Model cs.CV · 2025-02-14 · unverdicted · none · ref 176

  Step-Video-T2V describes a 30B-parameter text-to-video model with custom Video-VAE, 3D DiT, flow matching, and Video-DPO that claims state-of-the-art results on a new internal benchmark.

• Agent AI: Surveying the Horizons of Multimodal Interaction cs.AI · 2024-01-07 · unverdicted · none · ref 20

  The paper defines Agent AI as interactive multimodal systems that perceive grounded data and generate embodied actions, arguing this approach can mitigate hallucinations in foundation models.

• A Survey on Knowledge Distillation of Large Language Models cs.CL · 2024-02-20 · accept · none · ref 52

  A comprehensive survey of knowledge distillation for LLMs structured around algorithms, skill enhancement, and vertical applications, highlighting data augmentation as a key enabler.

• A Survey of Hallucination in Large Foundation Models cs.AI · 2023-09-12 · accept · none · ref 86

  A survey classifying hallucination phenomena specific to large foundation models, establishing evaluation criteria, examining mitigation strategies, and discussing future directions.

• Lessons from the Trenches on Reproducible Evaluation of Language Models cs.CL · 2024-05-23 · unreviewed · ref 121
• AppAgent: Multimodal Agents as Smartphone Users cs.CV · 2023-12-21 · unreviewed · ref 57