AIME 2025
PulseAugur coverage of AIME 2025 — every cluster mentioning AIME 2025 across labs, papers, and developer communities, ranked by signal.
6 day(s) with sentiment data
-
Adamas sparse attention boosts LLM long-context inference efficiency
Researchers have developed a new sparse attention mechanism called Adamas designed to improve the efficiency of long-context inference in large language models. Adamas utilizes the Hadamard transform, bucketing, and 2-b…
-
Kimi-K3.1-Pro-Coder model size slashed to 775 GB via multi-tier quantization
A new experimental quantization of the Kimi-K3.1-Pro-Coder model, named Kimi-K3.1-Pro-Coder (2.8T), has been developed, reducing its size from 1.77 TB to 775 GB. This was achieved through a multi-tiered quantization str…
-
New methods enhance LLM reasoning with automated step-wise rubrics · 2 sources tracked
Researchers have developed new methods for improving Large Language Model (LLM) reasoning through step-wise rubric rewards. The first approach, Step-wise Rubrics as Rewards (SRaR), uses an LLM judge to assign rubric ite…
-
New CIKA framework uses LLMs to pinpoint causal concepts for math reasoning
Researchers have developed CIKA, a novel framework designed to enhance Large Language Model (LLM) mathematical reasoning by identifying concepts that causally contribute to correct answers. Unlike previous methods that …
-
New framework streamlines formal proof generation for math problems
Researchers have developed a novel framework for generating formal proofs in systems like Lean 4, which integrates candidate answer selection and proof generation into a single process. This joint approach aims to reduc…
-
Metis-1.6 models achieve Pareto-optimal efficiency with novel architectures
A single researcher has developed Metis-1.6, a new suite of language models, Praxis and Logos, that prioritize efficient training methods and novel architectures over sheer parameter count. Despite not competing with fr…
-
New NormTune method boosts PEFT performance on reasoning tasks
Researchers have developed NormTune, a novel parameter-efficient fine-tuning (PEFT) strategy designed to improve performance on reasoning-intensive tasks. NormTune addresses the observed performance gap between PEFT and…
-
Adaptive Multi-Token Prediction accelerates LLM decoding up to 4.1x
Researchers have introduced Adaptive Multi-Token Prediction (AMTP), a novel method to accelerate language model decoding. AMTP dynamically adjusts the prediction horizon based on verification feedback, optimizing throug…
-
GigaChat 3.5 Reasoning benchmark claims scrutinized for misleading efficiency metrics
A recent analysis of AI benchmark tables highlights discrepancies in how performance claims are presented, using the GigaChat 3.5 Reasoning model and DeepSeek V4 Flash Preview as a case study. While GigaChat's published…
-
AI models tackle unsolved math problems with formal proofs and new discoveries
Two new research papers explore the capabilities of large language models in advanced mathematical reasoning and discovery. The first paper introduces Magenta, a system that bridges informal natural language math proble…
-
New research questions LLM aggregation strategies when candidates are wrong
A new research paper explores the effectiveness of different aggregation strategies for large language models, specifically when initial candidate answers are incorrect. The study, using the Qwen3-4B model on AIME-2025 …
-
New research explores test-time scaling for LLM reasoning
Two new research papers introduce novel methods for improving the reasoning capabilities of large language models (LLMs) through test-time scaling. The first paper, 'Consilience,' addresses limitations in existing confi…
-
New TTEL algorithm improves LLM reasoning efficiency by localizing errors
Researchers have developed a new inference-time algorithm called Test-Time Scaling via Error Localization (TTEL) to improve the efficiency of large language models on complex reasoning tasks. TTEL utilizes feedback to p…
-
KV-Cache Grafting boosts small LLMs, enabling 2.8M token context
Researchers have developed a novel technique called KV-Cache Grafting that enhances small language models without altering their weights. This method allows for the byte-exact restoration of verified knowledge into an i…
-
New ACPO framework enhances reinforcement learning for LLMs
Researchers have introduced Adaptive Credit Policy Optimization (ACPO), a new framework designed to improve credit assignment in reinforcement learning for large language models. ACPO addresses the challenge of sparse r…
-
TREK procedure boosts AI reasoning and agentic task performance
A new staged procedure called TREK (Teacher-Routed Exploration via Forward KL) has been introduced to improve the performance of AI models, particularly in complex reasoning tasks. TREK utilizes distillation not for dir…
-
TREK method boosts LLM reasoning by expanding exploration support
Researchers have introduced TREK (Teacher-Routed Exploration via Forward KL), a novel staged procedure designed to enhance the capabilities of language models, particularly in complex reasoning tasks. TREK utilizes dist…
-
New method uses wrong drafts to boost LLM math capabilities
Researchers have developed a novel technique called "Weak-to-Strong Elicitation via Mismatched Wrong Drafts" to improve the capabilities of large language models. This method involves using mathematically incorrect draf…
-
New framework VeryTrace verifies and repairs LLM reasoning traces
Researchers have developed VeryTrace, a new framework designed to verify and repair reasoning traces generated by large language models (LLMs). This system formalizes natural language reasoning into a structured, compil…
-
New benchmark TriggerBench reveals prospective memory challenges for LLMs
Researchers have introduced TriggerBench, a new benchmark designed to evaluate prospective memory (PM) in large language models (LLMs). Unlike retrospective memory (RM), which relies on explicit queries, PM assesses an …