Unboxing LLMs > loading...

Articles on Tech/AI/ML/LLMs

From Tokens to Tradeoffs.

Exploring Intelligence Through Language and Learning.

LLMs. Code. Theory. Clarity.

Thinking in Tokens: Essays on Language and Intelligence.

Navigating the LLM Revolution, One Shot at a Time.

Kimi K2: Engineering a Trillion‑Parameter Sparse MoE for Practical Agentic Intelligence

KV Cache Steering: A One‑Shot Route to Reasoning in Small LLMs

RLPR — Scaling Reinforcement Learning for LLM Reasoning Beyond Verifiers

Text‑to‑LoRA: Zero‑Friction Adaptation for LLMs

Magistral – Scaling Reasoning with Efficient GRPO Training

When LLMs Rewrite Their Own Rules: A Deep Dive into Self‑Adapting Language Models (SEAL)

Shrinking Reasoning LLMs with ‘R‑KV’

Darwin Gödel Machine: Engineering an Open‑Ended, Self‑Improving Coding Agent

Harnessing the Universal Geometry of Embeddings – A Vec2Vec Review

Self‑Evolving Agentic Workflows (SEW)

Panda: Patched Attention for Nonlinear Dynamics – A Deep‑Dive Review

SPECTRE: FFT-Powered Token Mixing for Long-Context Transformers

Qwen 3 – Reasoning-Native LLMS for the Real World

Self‑Learning at Absolute Zero: Reviewing the Absolute Zero Reasoner

“AReaL” – Decoupling Generation & Training

A Practical Guide to W8A8 Quantization – Running Large Language Models in (Much) Less Memory

From Scout to Behemoth: Inside Meta’s Llama 4 and the Race for 10-Million-Token Context

Agent S2 – Modular Intelligence for GUI Automation

Llama 4 and the 10 Million-Token Context: An Engineering Reality Check

Multi‑Token Attention – Moving Beyond Single‑Vector Focus

Measuring AI’s Growing “Attention Span”: A Review of METR Benchmark

Teaching Llama New Tricks: Memory-Efficient Trainable Tokens with LoRA

Aggressive Quantization: How Low Can You Go?

Four Habits of Highly Effective STaRs: Reviewing “Cognitive Behaviors that Enable Self‑Improving Reasoners”

Native Sparse Attention: Making Long‑Context Transformers Truly Practical

February 25, 2025

Structure Over Substance: A Deep Dive into “LLMs Can Easily Learn to Reason from Demonstrations”

February 23, 2025

AlphaGeometry 2: When the Machine Out-Reasons the Masters

February 21, 2025

Latent Depth, Infinite Thoughts: A Review of “Scaling up Test‑Time Compute with Latent Reasoning”

February 19, 2025

Decoding Distillation Scaling Laws : Compute‑Optimal Teacher–Student Training

February 9, 2025

“Thinking” LLMs on a Shoestring (S1): Fine-Tuning and Budget Forcing for Reasoning

February 2, 2025

Transformer^2 (Squared): Making Large Language Models Self‑Adaptive in Real Time

February 1, 2025

Teaching Small Gemma 2 Models to “Think” Like DeepSeek‑R1

January 21, 2025

DeepSeek‑R1’s Blueprint for Small Yet Smart LLMs : From Verifiable Rewards to Pocket‑Sized Reasoners

January 9, 2025

DeepSeek-V3 on a Shoestring: How I Run a 671B-parameter MoE Locally

December 23, 2024

Scaling the Wall: A Deep Dive into Best‑of‑N Jailbreaking

December 21, 2024

From Llama 3 to TÜLU 3: My Hands-On Recipe for SFT, DPO & RLVR

December 14, 2024

Coconut (Chain of Continuous Thought): Reasoning Beyond Tokens

December 10, 2024

AGORABENCH: Evaluating Language Models as Synthetic Data Generators

December 7, 2024

Reverse-Enhanced Thinking: Teaching Language Models to Reason Backwards

November 30, 2024

Pause Tokens for Transformers: Making Language Models *Pretend* to Think Before Speaking

November 5, 2024

Stealing User Prompts from Mixture‑of‑Experts: How Tiebreaks Become a Side‑Channel

November 2, 2024

Uncovering Mesa-Optimization in Transformers

October 31, 2024

Go Zero‑Shot: Faster & More Reproducible LLM Benchmarking

October 28, 2024

STRING: A Pragmatic Fix for LLMs’ Long‑Context Blind‑Spots

October 14, 2024

Buffer of Thoughts – Thought‑Augmented Reasoning for LLMs

September 28, 2024

Exploring Implicit Instruction Tuning: Instruction Following Even When You Don’t Try

September 21, 2024

Squeezing Big Models – K-Quantization + Importance Matrices for GGUF on a Plain CPU

September 12, 2024

Can LLMs Really Out‑Innovate Us? A Deep‑Dive Review of “Can LLMs Generate Novel Research Ideas?”

August 18, 2024

Fine-Tuning SmolLM on a Single GPU

August 13, 2024

4‑Bit Quantization Guide & Comparison Results – Running Llama 3.1 Efficiently

August 12, 2024

Evaluating the Evaluators: A Comprehensive Guide to LLM Benchmarks and Assessment Frameworks

Fine-Tuning Llama 3.1 70B on Consumer GPUs: A Practical Guide with FSDP and QLoRA

Shrinking Llama 3 with Intel AutoRound: Fast & Accurate Low-Bit Quantization

Mamba: Linear-Time Sequence Modeling Beyond Transformers

QLoRA Fine-Tuning Llama 3 with Axolotl on a Single GPU

Scaling Language Models Without Matrix Multiplication

MCT Self‑Refine: GPT‑4‑Calibre Math Reasoning With LLaMA‑3 8B

Binary Dreams: 1-bit & 2-bit Quantization with HQQ & Adapter Fine-Tuning

Teaching LLMs to Reinforce Themselves with Self‑Improving Robust Preference Optimization

Self‑Augmenting Llama 3: Stretching Models to 120B with Layer Surgery