Code and datasets

A pure Java implementation of OpenAI's gpt-oss inference in ~1000 lines of code optimized for CPU execution. Inspired by llama.cpp, llama2.c, this repo ports the gpt-oss PyTorch model.py to efficient Java code, emphasizing minimalism, simplicity, and educational purpose. Key features Pure Java - No native dependencies, runs anywhere with Java 23+ Complete gpt-oss architecture - Full implementation of MoE

TISER incorporates a multi-stage inference pipeline that combines explicit reasoning, timeline construction, and iterative self-reflection. The key idea behind our approach is to empower LLMs to adapt by scaling their internal reasoning process during inference. TISER enables models to systematically organize temporal information, verify their inferences, and refine their outputs.

CodeAssistBench provides a complete framework for evaluating AI coding assistants through multi-agent workflows, Docker validation, and comprehensive metrics collection. Built for researchers, ML engineers, and developers who need robust, reproducible evaluation of code generation agents.

Which set of features was responsible for a certain output of a machine learning model? Which components caused the failure of a cloud computing application? These are just two examples of questions we are addressing in this work by Identifying Coalition-based Explanations for Common and Rare Events in Any Model (ICECREAM). Specifically, we propose an information-theoretic quantitative measure for the influence

AgentOccam offers a simple but strong baseline for LLM-based web agents. By providing a URL and the task you want it to perform, AgentOccam can execute it for you. Its simplicity and effectiveness allow you to run it directly, or adapt it into a larger pipeline for its executing web tasks, such as web information retrieval before processing the documents. Without using in-context examples, new agent roles

A popular approach to building agents using Language Models (LMs) involves iteratively prompting the LM, reflecting on its outputs, and updating the input prompts until the desired task is achieved. However, our analysis reveals two key shortcomings in the existing methods: (i) limited exploration of the decision space due to repetitive reflections, which result in redundant inputs, and (ii) an inability

The effectiveness of automatic evaluation of generative models is typically measured by comparing the labels generated via automation with human labels using correlation metrics. However, metrics like Krippendorff’s α and Randolph’s κ were originally designed to measure the reliability of human labeling, thus make assumptions about typical human labeling behavior, and these assumptions may not be applicable

Open domain question answering systems frequently rely on information retrieved from large collections of text (such as the Web) to answer questions. However, such collections of text often contain conflicting information, and indiscriminately depending on this information may result in untruthful and inaccurate answers. To understand the gravity of this problem, we collect a human-annotated dataset, Question

We present ECI-sampling (stands for extrapolation-correction-interpolation) as a novel framework for adapting pre-trained, unconstrained generative models to exactly satisfy constraints in a zero-shot manner, without requiring expensive gradient computations or fine-tuning.

Can integrating spectral and curvature signals unlock new potential in graph representation learning? Non-Euclidean geometries, particularly Riemannian manifolds such as hyperbolic (negative curvature) and spherical (positive curvature), offer powerful inductive biases for embedding complex graph structures like scale-free, hierarchical, and cyclic patterns. Meanwhile, spectral filtering excels at processing

This repository will contain the official implementation of the NAACL'2025 paper Constrained Decoding with Speculative Lookaheads

Recent advancements in language-guided diffusion models for image editing are often bottle-necked by cumbersome prompt engineering to precisely articulate desired changes. An intuitive alternative calls on guidance from in-the-wild image exemplars to help users bring their imagined edits to life. Contemporary exemplar-based editing methods shy away from leveraging the rich latent space learnt by pre-existing