Large Language Models

I am not an "anti" like this guy, but still an interesting video of person interacting with chat 4o

(Posting Here because removed by Chatgpt Complaints moderators because the model here is 4o, and refuse to believe there were any safety ...

We built a way for two people's AI context to talk to each other (without sharing their conversations)

We've been thinking about how we use AI in our relationships. Big part of it is about other people. Talking about them, figuring out what...

No flattery please, Claude: I’m British | Brief letters

[2510.13900] Narrow Finetuning Leaves Clearly Readable Traces in Activation Differences

Abstract page for arXiv paper 2510.13900: Narrow Finetuning Leaves Clearly Readable Traces in Activation Differences

[2510.13315] Self-Aug: Query and Entropy Adaptive Decoding for Large Vision-Language Models

Abstract page for arXiv paper 2510.13315: Self-Aug: Query and Entropy Adaptive Decoding for Large Vision-Language Models

[2510.06084] Spectrum Tuning: Post-Training for Distributional Coverage and In-Context Steerability

Abstract page for arXiv paper 2510.06084: Spectrum Tuning: Post-Training for Distributional Coverage and In-Context Steerability

[2509.22641] Death of the Novel(ty): Beyond n-Gram Novelty as a Metric for Textual Creativity

Abstract page for arXiv paper 2509.22641: Death of the Novel(ty): Beyond n-Gram Novelty as a Metric for Textual Creativity

[2509.21091] Best-of-$\infty$ -- Asymptotic Performance of Test-Time Compute

Abstract page for arXiv paper 2509.21091: Best-of-$\infty$ -- Asymptotic Performance of Test-Time Compute

[2509.20986] SiNGER: A Clearer Voice Distills Vision Transformers Further

Abstract page for arXiv paper 2509.20986: SiNGER: A Clearer Voice Distills Vision Transformers Further

[2509.12610] ScaleDoc: Scaling LLM-based Predicates over Large Document Collections

Abstract page for arXiv paper 2509.12610: ScaleDoc: Scaling LLM-based Predicates over Large Document Collections

[2509.10625] No Answer Needed: Predicting LLM Answer Accuracy from Question-Only Linear Probes

Abstract page for arXiv paper 2509.10625: No Answer Needed: Predicting LLM Answer Accuracy from Question-Only Linear Probes

[2509.05425] No Text Needed: Forecasting MT Quality and Inequity from Fertility and Metadata

Abstract page for arXiv paper 2509.05425: No Text Needed: Forecasting MT Quality and Inequity from Fertility and Metadata

[2511.10833] SURFACEBENCH: A Geometry-Aware Benchmark for Symbolic Surface Discovery

Abstract page for arXiv paper 2511.10833: SURFACEBENCH: A Geometry-Aware Benchmark for Symbolic Surface Discovery

[2511.08939] TransactionGPT

Abstract page for arXiv paper 2511.08939: TransactionGPT

[2507.05890] Psychometric Item Validation Using Virtual Respondents with Trait-Response Mediators

Abstract page for arXiv paper 2507.05890: Psychometric Item Validation Using Virtual Respondents with Trait-Response Mediators

[2507.01335] LEDOM: Reverse Language Model

Abstract page for arXiv paper 2507.01335: LEDOM: Reverse Language Model

[2510.15165] Policy Transfer for Continuous-Time Reinforcement Learning: A (Rough) Differential Equation Approach

Abstract page for arXiv paper 2510.15165: Policy Transfer for Continuous-Time Reinforcement Learning: A (Rough) Differential Equation App...

[2506.17871] LLM Probability Concentration: How Alignment Shrinks the Generative Horizon

Abstract page for arXiv paper 2506.17871: LLM Probability Concentration: How Alignment Shrinks the Generative Horizon

[2510.10902] Auditing Information Disclosure During LLM-Scale Gradient Descent Using Gradient Uniqueness

Abstract page for arXiv paper 2510.10902: Auditing Information Disclosure During LLM-Scale Gradient Descent Using Gradient Uniqueness

[2510.04573] LaDiR: Latent Diffusion Enhances LLMs for Text Reasoning

Abstract page for arXiv paper 2510.04573: LaDiR: Latent Diffusion Enhances LLMs for Text Reasoning

[2510.08646] Mitigating Over-Refusal in Aligned Large Language Models via Inference-Time Activation Energy

Abstract page for arXiv paper 2510.08646: Mitigating Over-Refusal in Aligned Large Language Models via Inference-Time Activation Energy

[2506.11103] You Only Fine-tune Once: Many-Shot In-Context Fine-Tuning for Large Language Models

Abstract page for arXiv paper 2506.11103: You Only Fine-tune Once: Many-Shot In-Context Fine-Tuning for Large Language Models

[2509.23202] Bridging the Gap Between Promise and Performance for Microscaling FP4 Quantization

Abstract page for arXiv paper 2509.23202: Bridging the Gap Between Promise and Performance for Microscaling FP4 Quantization