AI Papers of the Week
Every paper worth reading in AI, hand-picked one week at a time.
ALMA
ALMA (Automated meta-Learning of Memory designs for Agentic systems) from Jeff Clune’s group introduces a Meta Agent that automatically discovers memory designs for agentic systems through open-ended exploration in code space. Instead of relying on hand-engineered memory modules, ALMA searches over database schemas, retrieval mechanisms, and update strategies expressed as executable code, consistently outperforming all human-designed memory baselines across four sequential decision-making benchmarks.
LLaDA 2.1
Ant Group releases LLaDA 2.1, a major upgrade to discrete diffusion language models that breaks the speed-quality trade-off through Token-to-Token (T2T) editing. By weaving token editing into the conventional Mask-to-Token decoding scheme, LLaDA 2.1 introduces two configurable modes: Speedy Mode for aggressive throughput and Quality Mode for benchmark-leading accuracy. The release also includes the first large-scale RL framework for diffusion LLMs.
SkillRL
SkillRL introduces a recursive skill-augmented RL framework that bridges the gap between raw experience and policy improvement through automatic skill discovery. Instead of storing noisy raw trajectories, SkillRL distills experience into reusable high-level behavioral patterns and evolves them alongside the agent policy during training.
InftyThink+
InftyThink+ is an end-to-end RL framework for infinite-horizon reasoning that optimizes the entire iterative reasoning trajectory. Standard long chain-of-thought suffers from quadratic cost, context length limits, and lost-in-the-middle degradation. InftyThink+ addresses all three by letting models autonomously decide when to summarize, what to preserve, and how to resume, trained through trajectory-level reinforcement learning.
Agyn
Agyn is a fully automated multi-agent system that models software engineering as an organizational process rather than a monolithic code generation task. Built on an open-source platform for configuring agent teams, the system assigns specialized agents to distinct roles and follows a structured development methodology - all without human intervention. Notably, Agyn was designed for real production use and was not tuned for the SWE-bench.
AdaptEvolve
AdaptEvolve tackles a key efficiency bottleneck in evolutionary agentic systems: the repeated invocation of large LLMs during iterative refinement loops. The method uses intrinsic generation confidence to dynamically select which model to invoke at each step, routing easy sub-problems to smaller models and reserving expensive frontier models for genuinely hard decisions.
Gaia2
Meta FAIR introduces Gaia2, a next-generation agent benchmark where environments change independently of agent actions, forcing agents to handle temporal pressure, uncertainty, and multi-agent coordination. GPT-5 leads at 42% pass@1 but struggles with time-constrained tasks, while Kimi-K2 leads open-source models at 21%. Built on the open-source Agents Research Environments (ARE) platform with action-level verifiers, Gaia2 represents a paradigm shift from static benchmarks to dynamic evaluation of agentic capabilities.
AgentArk
AgentArk distills multi-agent debate dynamics into a single LLM, transferring the reasoning and self-correction abilities of multi-agent systems into one model at training time. Three hierarchical distillation strategies (reasoning-enhanced SFT, trajectory-based augmentation, and process-aware distillation with a process reward model) yield an average 4.8% improvement over single-agent baselines across math and reasoning benchmarks, approaching full multi-agent performance at a fraction of the inference cost. Cross-family distillation (e.g., Qwen3-32B to LLaMA-3-8B) produces the largest gains, suggesting heterogeneous architectures benefit most from transferred reasoning signals.
Semi-Autonomous Mathematics Discovery with Gemini
This paper from Google DeepMind presents a case study in semi-autonomous mathematics discovery using Aletheia, a specialized math research agent built on Gemini Deep Think. The team systematically evaluated 700 open conjectures from Bloom’s Erdos Problems database, combining AI-driven natural language verification with human expert evaluation, and addressed 13 previously open problems.
TinyLoRA
This paper from Meta FAIR asks how small a LoRA adapter can get and still teach a model to reason. The answer: remarkably small. The authors propose TinyLoRA, a method that scales low-rank adapters down to as few as one trainable parameter by projecting through fixed random tensors and sharing weights across all modules. The key insight is that RL makes fundamentally more information-dense updates than SFT, enabling effective learning with orders of magnitude fewer parameters.
xMemory
xMemory argues that standard RAG retrieval is a poor fit for agent memory because the evidence source is a bounded, coherent dialogue stream where candidate spans are highly correlated near-duplicates. Fixed top-k similarity retrieval collapses into a single dense region, returning redundant context, while post-hoc pruning can break temporally linked evidence chains. xMemory replaces this with hierarchical memory construction and structure-aware top-down retrieval.
InfMem
InfMem is a cognitive agent for ultra-long document QA that uses System-2-style control to actively manage bounded memory. Instead of passively compressing each chunk as it streams in, InfMem runs a PreThink-Retrieve-Write loop that monitors evidence sufficiency, fetches missing facts from anywhere in the document, and compresses everything into a fixed-size memory - then stops early once it has enough.
A-RAG
A-RAG is an agentic RAG framework that gives LLMs direct access to hierarchical retrieval interfaces instead of relying on fixed retrieval algorithms or predefined workflows. The agent autonomously decides what to search, at which granularity, and when to stop - representing a paradigm shift from static retrieval pipelines to truly agentic information gathering.
Agent Primitives
Agent Primitives introduces reusable latent building blocks for LLM-based multi-agent systems. Inspired by how neural networks are built from composable modules like residual blocks and attention heads, the authors decompose existing MAS architectures into three recurring computation patterns that communicate via KV cache instead of natural language, reducing error accumulation and boosting efficiency.
OpenScholar
OpenScholar is a fully open, retrieval-augmented language model designed for scientific literature synthesis. It retrieves passages from a datastore of 45 million open-access papers, generates citation-backed responses, and iteratively refines outputs through a self-feedback loop. On ScholarQABench, the first large-scale multi-domain benchmark for literature search, OpenScholar-8B outperforms GPT-4o by 6% and PaperQA2 by 5.5% in correctness, while achieving citation accuracy on par with human experts and being preferred over expert-written answers 51-70% of the time.
Kimi K2.5: Visual Agentic Intelligence
Kimi K2.5 is an open-source multimodal agentic model from Moonshot AI that jointly optimizes text and vision capabilities through native multimodal pretraining on 15 trillion mixed tokens, zero-vision SFT, and joint reinforcement learning. K2.5 also introduces Agent Swarm, a parallel agent orchestration framework that dynamically decomposes complex tasks into concurrent subtasks, reducing latency by up to 4.5x over single-agent baselines.
Shaping Capabilities with Token-Level Data Filtering
Researchers from Anthropic and Stanford show that filtering pretraining data at the token level is a highly effective, scalable, and robust approach for selectively removing undesired capabilities from language models. Using medical knowledge removal as a proxy task, token-level filtering Pareto dominates document-level filtering and achieves a 7,000x compute slowdown on the target domain for 1.8B parameter models - while preserving capabilities in related fields.
How AI Impacts Skill Formation
Researchers from Anthropic conducted randomized experiments to study how AI assistance affects the development of software engineering skills. They find that using AI to complete coding tasks with a new Python library significantly impaired conceptual understanding, code reading, and debugging abilities - without delivering significant efficiency gains on average.
Reinforcement Learning via Self-Distillation
This paper introduces Self-Distillation Policy Optimization (SDPO), an on-policy RL algorithm that converts rich textual feedback from verifiable environments into dense credit assignment without requiring an external teacher model. SDPO uses the current model conditioned on feedback as a “self-teacher” to retrospectively identify mistakes in its own rollouts, substantially outperforming GRPO across scientific reasoning, tool use, and competitive programming.
Self-Improving Pretraining
Self-Improving Pretraining is a new pretraining paradigm from Meta FAIR that replaces standard next-token prediction with sequence-level generation guided by an existing post-trained model acting as both a suffix rewriter and a suffix judge. The approach addresses quality, safety, and factuality issues at pretraining time rather than deferring them to post-training, yielding large gains across all three dimensions.
LingBot-World: Open-Source World Simulator
LingBot-World is an open-source world simulator that evolves a video generation model into an interactive, real-time environment engine. Built on a 28B-parameter Mixture-of-Experts architecture, it achieves high-fidelity dynamics across diverse domains with sub-second latency at 16 fps, outperforming Genie 3 and Mirage 2 in dynamic degree while being fully open-source.
TTT-Discover: Learning to Discover at Test Time
TTT-Discover introduces test-time training for scientific discovery, performing reinforcement learning at test time so the LLM can continue to train with experience specific to the test problem. Unlike prior work like AlphaEvolve that prompts a frozen LLM, this approach enables the model itself to improve while attempting to solve hard problems.
Memory Control for Long-Horizon Agents
This paper introduces the Agent Cognitive Compressor (ACC), a bio-inspired mechanism that addresses degraded agent behavior in long multi-turn workflows caused by loss of constraint focus, error accumulation, and memory-induced drift. ACC replaces continuous transcript retention with a bounded internal state that updates incrementally during each interaction turn.
Benchmarking Agents on Hard CLI Tasks
Terminal-Bench 2.0 presents a carefully curated hard benchmark composed of 89 tasks in computer terminal environments inspired by problems from real workflows. Each task features a unique environment, human-written solution, and comprehensive tests for verification, addressing the gap where current benchmarks either don’t measure real-world tasks or aren’t sufficiently difficult.