Skip to content
Navigation menu
Search
Powered by Algolia
Search
Log in
Create account
DEV Community
Close
Measuring ML systems honestly Series' Articles
Back to Raihan's Series
How Accurate Is Your Model Right Now? Estimating Accuracy Without Labels
Raihan
Raihan
Raihan
Follow
Oct 4
How Accurate Is Your Model Right Now? Estimating Accuracy Without Labels
#
machinelearning
#
mlops
#
python
#
monitoring
1
reaction
Comments
7
comments
4 min read
Moderation
AWQ Looked 10 Points Better on GSM8K Until I Stopped Truncating the Answers
Raihan
Raihan
Raihan
Follow
Oct 5
AWQ Looked 10 Points Better on GSM8K Until I Stopped Truncating the Answers
#
llm
#
machinelearning
#
python
#
testing
Comments
4
comments
12 min read
Moderation
Do LLMs Invent Japanese Law Articles? A Bilingual Benchmark
Raihan
Raihan
Raihan
Follow
Oct 6
Do LLMs Invent Japanese Law Articles? A Bilingual Benchmark
#
llm
#
nlp
#
japanese
#
opensource
Comments
Add Comment
4 min read
Moderation
Small LLM Judges Approved 11% and 41% of Wrong Answers. Then I Fixed My Own Pairwise Test.
Raihan
Raihan
Raihan
Follow
Oct 7
Small LLM Judges Approved 11% and 41% of Wrong Answers. Then I Fixed My Own Pairwise Test.
#
llm
#
machinelearning
#
python
#
evaluation
3
reactions
Comments
6
comments
5 min read
Moderation
GraphProof-QA: Teaching a Small Model to Prove Its Answers
Raihan
Raihan
Raihan
Follow
Oct 8
GraphProof-QA: Teaching a Small Model to Prove Its Answers
#
llm
#
machinelearning
#
python
#
nlp
1
reaction
Comments
2
comments
4 min read
Moderation
Mentioned Is Not Binding: Which Solicitation Clauses Actually Bind the Contract
Raihan
Raihan
Raihan
Follow
Oct 8
Mentioned Is Not Binding: Which Solicitation Clauses Actually Bind the Contract
#
llm
#
machinelearning
#
python
#
nlp
Comments
Add Comment
3 min read
Moderation
A Grammar Stopped My Model Inventing FAR Clauses. It Also Made It Refuse the Real Ones
Raihan
Raihan
Raihan
Follow
Oct 9
A Grammar Stopped My Model Inventing FAR Clauses. It Also Made It Refuse the Real Ones
#
llm
#
machinelearning
#
python
#
nlp
Comments
2
comments
4 min read
Moderation
I Distilled a 568M Multilingual Model Into a 37M Japanese-English Encoder — Here's What Survived
Raihan
Raihan
Raihan
Follow
Oct 10
I Distilled a 568M Multilingual Model Into a 37M Japanese-English Encoder — Here's What Survived
#
machinelearning
#
nlp
#
python
#
japanese
1
reaction
Comments
Add Comment
4 min read
Moderation
Teaching a Computer to Hear Japanese Pitch Accent
Raihan
Raihan
Raihan
Follow
Oct 10
Teaching a Computer to Hear Japanese Pitch Accent
#
llm
#
machinelearning
#
python
#
japanese
1
reaction
Comments
Add Comment
2 min read
Moderation
My Triton Kernel Ties bitsandbytes — and Both Lose to Math
Raihan
Raihan
Raihan
Follow
Oct 12
My Triton Kernel Ties bitsandbytes — and Both Lose to Math
#
gpu
#
llm
#
python
#
performance
Comments
Add Comment
4 min read
Moderation
Was My 2,103-Token Tokenizer a Mistake?
Raihan
Raihan
Raihan
Follow
Oct 13
Was My 2,103-Token Tokenizer a Mistake?
#
machinelearning
#
llm
#
python
#
nlp
Comments
Add Comment
3 min read
Moderation
DemoDoctor: Bad Robot Demonstrations Are Detectable. Do They Matter?
Raihan
Raihan
Raihan
Follow
Oct 14
DemoDoctor: Bad Robot Demonstrations Are Detectable. Do They Matter?
#
machinelearning
#
robotics
#
python
#
statistics
1
reaction
Comments
1
comment
3 min read
Moderation
Invoice-Check JP: A 3B Model Reads the Invoice, Checks Decide Whether to Trust It
Raihan
Raihan
Raihan
Follow
Oct 15
Invoice-Check JP: A 3B Model Reads the Invoice, Checks Decide Whether to Trust It
#
machinelearning
#
llm
#
python
#
japanese
Comments
Add Comment
4 min read
Moderation
PO Matching Looked Like It Halved Unsafe Invoice Approvals Until I Checked Why
Raihan
Raihan
Raihan
Follow
Oct 16
PO Matching Looked Like It Halved Unsafe Invoice Approvals Until I Checked Why
#
langgraph
#
llm
#
python
#
evaluation
1
reaction
Comments
1
comment
5 min read
Moderation
I Pre-Registered a Four-Way LangGraph Shootout. Plan-and-Execute Cost 2.3x and Bought Nothing I Could Detect
Raihan
Raihan
Raihan
Follow
Oct 17
I Pre-Registered a Four-Way LangGraph Shootout. Plan-and-Execute Cost 2.3x and Bought Nothing I Could Detect
#
langgraph
#
llm
#
python
#
evaluation
1
reaction
Comments
1
comment
6 min read
Moderation
Edit the Weights, Retrieve the Fact, or Look It Up in a Graph?
Raihan
Raihan
Raihan
Follow
Oct 18
Edit the Weights, Retrieve the Fact, or Look It Up in a Graph?
#
llm
#
machinelearning
#
python
#
evaluation
Comments
1
comment
2 min read
Moderation
Your LLM Judge Is Lying to You, a Little
Raihan
Raihan
Raihan
Follow
Oct 19
Your LLM Judge Is Lying to You, a Little
#
llm
#
machinelearning
#
python
#
statistics
2
reactions
Comments
3
comments
2 min read
Moderation
The Verb Comes Last: Measuring Why Japanese-English Live Translation Waits
Raihan
Raihan
Raihan
Follow
Oct 20
The Verb Comes Last: Measuring Why Japanese-English Live Translation Waits
#
machinelearning
#
python
#
nlp
#
speech
Comments
Add Comment
3 min read
Moderation
My Japanese Tutor Agent Works — Except Its Brain Calls Everything a Particle
Raihan
Raihan
Raihan
Follow
Oct 21
My Japanese Tutor Agent Works — Except Its Brain Calls Everything a Particle
#
machinelearning
#
python
#
nlp
#
langchain
Comments
Add Comment
3 min read
Moderation
Do Japanese and English Share Neurons? I Trained the Microscope to Check
Raihan
Raihan
Raihan
Follow
Oct 22
Do Japanese and English Share Neurons? I Trained the Microscope to Check
#
machinelearning
#
python
#
nlp
#
llm
Comments
Add Comment
2 min read
Moderation
One Knob, 19x Latency: What I Learned Benchmarking vLLM on a Single GPU
Raihan
Raihan
Raihan
Follow
Oct 23
One Knob, 19x Latency: What I Learned Benchmarking vLLM on a Single GPU
#
machinelearning
#
python
#
llm
#
devops
4
reactions
Comments
4
comments
2 min read
Moderation
Correct, Cited, and Two Years Out of Date: Measuring Version Leakage in RAG
Raihan
Raihan
Raihan
Follow
Oct 24
Correct, Cited, and Two Years Out of Date: Measuring Version Leakage in RAG
#
machinelearning
#
python
#
nlp
#
llm
1
reaction
Comments
Add Comment
2 min read
Moderation
Your RAG Has No Idea What Year It Is: A Frozen Benchmark for Version Leakage
Raihan
Raihan
Raihan
Follow
Oct 25
Your RAG Has No Idea What Year It Is: A Frozen Benchmark for Version Leakage
#
machinelearning
#
python
#
nlp
#
llm
Comments
Add Comment
2 min read
Moderation
Your MCP Tools Are Leaking: A Benchmark for Description Canonicalisation
Raihan
Raihan
Raihan
Follow
Oct 26
Your MCP Tools Are Leaking: A Benchmark for Description Canonicalisation
#
machinelearning
#
python
#
security
#
llm
1
reaction
Comments
2
comments
3 min read
Moderation
We're a place where coders share, stay up-to-date and grow their careers.
Log in
Create account