AI Papers of the Week
Every paper worth reading in AI, hand-picked one week at a time.
OdysseyBench
A new benchmark and data‑generation pipeline to test agents on realistic, multi‑day office tasks across Word, Excel, PDF, Email, and Calendar. It introduces OdysseyBench (two splits) and HOMERAGENTS (an automated multi‑agent generator), with evaluations showing large gaps between human and model performance and clear benefits from semantically compressed memory.
Beyond Ten Turns
This paper introduces ASearcher, an open-source framework for training LLM-based search agents capable of long-horizon, expert-level search. It addresses two major limitations in prior open-source approaches: short turn limits (≤10) and lack of large-scale, high-quality QA data.
GLM-4.5
An open Mixture‑of‑Experts family that targets a single model excelling across agentic tool use, complex reasoning, and real‑world coding. GLM‑4.5 (355B total, 32B active) introduces hybrid “thinking vs direct” modes, multi‑stage pretrain + mid‑train to 128K context, and extensive RL for reasoning, agents, and instruction following. It ranks near the top on a 12‑bench ARC suite and releases weights and eval tooling.
A Deep Dive into RL for LLM Reasoning
This paper reviews and rigorously re-evaluates reinforcement learning techniques for LLM reasoning, addressing inconsistencies caused by varied setups and unclear guidelines. It offers a unified open-source framework, practical selection guidelines, and shows that a minimalist two-technique combo with vanilla PPO can outperform methods like GRPO and DAPO.
Is Chain-of-Thought Reasoning a Mirage?
Researchers from Arizona State University investigate whether Chain-of-Thought (CoT) reasoning in LLMs reflects genuine logical inference or mere pattern replication from training data. They introduce a data distribution lens to analyze CoT’s dependence on in-distribution patterns and its brittleness under distribution shifts.
Agentic Web
This paper introduces the concept of the Agentic Web, a transformative vision of the internet where autonomous AI agents, powered by LLMs, act on behalf of users to plan, coordinate, and execute tasks. It proposes a structured framework for understanding this shift, situating it as a successor to the PC and Mobile Web eras. The Agentic Web is defined by a triplet of core dimensions, intelligence, interaction, and economics, and involves fundamental architectural and commercial transitions.
ReaGAN
This paper introduces ReaGAN, a graph learning framework that reconceptualizes each node in a graph as an autonomous agent capable of planning, reasoning, and acting via a frozen LLM. Instead of relying on static, layer-wise message passing, ReaGAN enables node-level autonomy, where each node independently decides whether to aggregate information from local neighbors, retrieve semantically similar but distant nodes, or take no action at all. This node-agent abstraction addresses two key challenges in graph learning: (1) handling varying informativeness of nodes and (2) combining local structure with global semantics.
Tool-Augmented Unified Retrieval Agent for AI Search
Presents a production-ready framework that extends the RAG (Retrieval-Augmented Generation) paradigm to support real-time, dynamic, and transactional queries through agentic tool use. Unlike conventional RAG systems that rely on static web snapshots, TURA enables LLM-based systems to interact with external APIs and databases, addressing user intents that require up-to-date or structured information (e.g., train schedules, weather forecasts).
Tabular Data Understanding with LLMs
This survey reviews LLM and MLLM methods for table understanding, outlining a taxonomy of tabular representations and tasks. It identifies key gaps, including limited reasoning beyond retrieval, difficulties with complex or large-scale tables, and poor generalization across diverse formats.
Medical Reasoning in the Era of LLMs
This review categorizes techniques for enhancing LLM medical reasoning into training-time (e.g., fine-tuning, RL) and test-time (e.g., prompt engineering, multi-agent systems) approaches, applied across modalities and clinical tasks. It highlights advances in evaluation methods, key challenges like the faithfulness–plausibility gap, and the need for native multimodal reasoning in future medical AI.
Geometric-Mean Policy Optimization
Introduces a stabilized alternative to Group Relative Policy Optimization (GRPO), which is widely used to improve reasoning capabilities in large language models via reinforcement learning. GRPO optimizes the arithmetic mean of token-level rewards but suffers from training instability due to extreme importance sampling ratios. GMPO addresses this by instead maximizing the geometric mean of token-level rewards, leading to more stable updates.
Graph-R1
Introduces a novel RAG framework that moves beyond traditional one-shot or chunk-based retrieval by integrating graph-structured knowledge, agentic multi-turn interaction, and RL. The core goal is to improve factual accuracy, retrieval efficiency, and reasoning quality in knowledge-intensive tasks.
Hierarchical Reasoning Model
Hierarchical Reasoning Model (HRM) is a novel, brain-inspired architecture that replaces CoT prompting with a recurrent model designed for deep, latent computation. It departs from token-level reasoning by using two coupled modules: a slow, high-level planner and a fast, low-level executor, achieving greater reasoning depth and efficiency with only 27M parameters and no pretraining. Despite its small size and minimal training data (~1k examples), HRM solves complex tasks like ARC, Sudoku-Extreme, and 30×30 maze navigation, where CoT-based LLMs fail.
Where to show demos in your prompt?
Introduces DPP bias, a new kind of positional sensitivity in LLM where the location of demonstrations in a prompt significantly affects output accuracy and stability. While prior work focused on demo content and order, this study reveals that moving an identical demo block across different sections of a prompt, e.g., before vs. after the user query, can change accuracy by up to 20 points and flip a large percentage of model predictions.
Subliminal Learning
This paper introduces and analyzes a phenomenon the authors term subliminal learning: the transfer of behavioral traits between language models through semantically unrelated training data. Specifically, when a teacher model exhibiting a trait (e.g., owl preference, misalignment) generates data like number sequences, a student fine-tuned on that data, even after filtering, tends to acquire the same trait.
Gemini 2.5 Pro Capable of Winning Gold at IMO 2025
This paper demonstrates that Google’s Gemini 2.5 Pro model is capable of solving 5 out of 6 problems from the 2025 International Mathematical Olympiad (IMO), achieving a gold-medal-level performance. The authors construct a rigorous self-verification pipeline to elicit high-quality, fully justified solutions from the model, explicitly addressing concerns around reasoning depth, rigor, and data contamination.
Inverse Scaling in Test-Time Compute
This paper presents a systematic study of inverse scaling in large reasoning models (LRMs), where increasing the test-time compute (i.e., reasoning length) harms rather than helps model performance. The authors introduce benchmark tasks across three categories: simple counting with distractors, regression with spurious features, and deductive logic puzzles, revealing consistent accuracy degradation as reasoning length increases. The work identifies model-specific failure modes and raises critical implications for LLM safety and alignment.
Deep Researcher with Test-Time Diffusion
Rethinks how deep research agents generate long-form reports. Rather than relying on static inference strategies like Chain-of-Thought or best-of-n sampling, TTD-DR frames the report generation process as a diffusion process. It starts with a noisy draft and iteratively refines it through retrieval-enhanced denoising, guided by a structured plan. This iterative loop mimics how human researchers search, reason, revise, and accumulate context over time.
One Token to Fool LLM-as-a-Judge
Investigates the surprising fragility of LLM-based reward models used in Reinforcement Learning with Verifiable Rewards (RLVR). The authors find that inserting superficial, semantically empty tokens, like “Thought process:”, “Solution”, or even just a colon “:”, can consistently trick models into giving false positive rewards, regardless of the actual correctness of the response.
Agentic-R1
This paper introduces Agentic-R1, a 7B language model trained to dynamically switch between tool-based execution and pure text reasoning using a novel fine-tuning framework called DualDistill. Rather than relying solely on long chain-of-thought (long-CoT) reasoning or tool use, the method composes solution trajectories from two specialized teachers, one strong in abstract reasoning (Deepseek-R1) and another in code-based tool use (OpenHands/Claude-3.5). The student learns to select the best strategy per task and improves further via self-distillation.
Chain-of-Thought Monitorability
Proposes that language-based CoT reasoning in LLMs offers an opportunity for AI safety by enabling automated oversight of models’ internal reasoning processes. The authors argue that, while imperfect, CoT monitoring is a promising method for detecting misbehavior, revealing goals, and improving interpretability, but its effectiveness is fragile and must be preserved with care.
Stress Testing Large Reasoning Models
Proposes a new benchmark framework called REST to evaluate the robustness of Large Reasoning Models (LRMs) under multi-question stress. Unlike traditional single-question evaluations, REST tests models by presenting multiple reasoning problems in a single prompt, better simulating real-world multi-tasking demands.
Scaling up RL
This paper investigates how prolonged RL can enhance reasoning abilities in small language models across diverse domains. Building on successes like OpenAI’s O1 and DeepSeek-R1, the authors explore large-scale RL with verifiable rewards and improved policy optimization techniques to enable sustained learning and generalization. They introduce the Nemotron-Research-Reasoning-Qwen-1.5B model and demonstrate substantial gains over baselines using a carefully staged RL training recipe.
Machine Bullshit
This paper introduces the concept of machine bullshit, extending Harry Frankfurt’s definition, discourse made with indifference to truth, to LLMs. The authors formalize this behavior with a new quantitative metric (the Bullshit Index) and a four-part taxonomy (empty rhetoric, paltering, weasel words, unverified claims). Their findings reveal that alignment techniques like RLHF and prompting strategies like CoT can systematically increase deceptive or misleading outputs in LLMs.