Oh wow!
A surprising result from Databricks when measuring embeddings and rerankers on internal evals.
1- Reranking few docs improves recall (expected).
2- Reranking many docs degrades quality (!).
3- Reranking too many documents is quite often worse than using embedding model alone (!!).
A surprising result from Databricks when measuring embeddings and rerankers on internal evals.
1- Reranking few docs improves recall (expected).
2- Reranking many docs degrades quality (!).
3- Reranking too many documents is quite often worse than using embedding model alone (!!).
November 20, 2024 at 8:46 PM
Oh wow!
A surprising result from Databricks when measuring embeddings and rerankers on internal evals.
1- Reranking few docs improves recall (expected).
2- Reranking many docs degrades quality (!).
3- Reranking too many documents is quite often worse than using embedding model alone (!!).
A surprising result from Databricks when measuring embeddings and rerankers on internal evals.
1- Reranking few docs improves recall (expected).
2- Reranking many docs degrades quality (!).
3- Reranking too many documents is quite often worse than using embedding model alone (!!).
hikes! ranking donuts! reranking parks with new playgrounds!
truly the pinnacle of an early fall weekend
truly the pinnacle of an early fall weekend
October 7, 2023 at 11:32 PM
hikes! ranking donuts! reranking parks with new playgrounds!
truly the pinnacle of an early fall weekend
truly the pinnacle of an early fall weekend
Our new article in @science.org enables social media reranking outside of platforms' walled gardens.
We add an LLM-powered reranking of highly polarizing political content into N=1256 participants' feeds. Downranking cools tensions with the opposite party—but upranking inflames them.
We add an LLM-powered reranking of highly polarizing political content into N=1256 participants' feeds. Downranking cools tensions with the opposite party—but upranking inflames them.
December 1, 2025 at 7:33 PM
Our new article in @science.org enables social media reranking outside of platforms' walled gardens.
We add an LLM-powered reranking of highly polarizing political content into N=1256 participants' feeds. Downranking cools tensions with the opposite party—but upranking inflames them.
We add an LLM-powered reranking of highly polarizing political content into N=1256 participants' feeds. Downranking cools tensions with the opposite party—but upranking inflames them.
Realtime LLM social media reranking enables field experiments demonstrating how we can depolarize. Congratulations @tiziano.bsky.social, @msaveski.bsky.social, and @jiachenyan.bsky.social!
New paper: Do social media algorithms shape affective polarization?
We ran a field experiment on X/Twitter (N=1,256) using LLMs to rerank content in real-time, adjusting exposure to polarizing posts. Result: Algorithmic ranking impacts feelings toward the political outgroup! 🧵⬇️
We ran a field experiment on X/Twitter (N=1,256) using LLMs to rerank content in real-time, adjusting exposure to polarizing posts. Result: Algorithmic ranking impacts feelings toward the political outgroup! 🧵⬇️
November 26, 2024 at 12:08 AM
Realtime LLM social media reranking enables field experiments demonstrating how we can depolarize. Congratulations @tiziano.bsky.social, @msaveski.bsky.social, and @jiachenyan.bsky.social!
Maybe a little cash will speed up the next reranking...
September 27, 2026 at 1:12 PM
Maybe a little cash will speed up the next reranking...
Mat is not on 🦋—posting on his behalf!
It's time to revisit common assumptions in IR! Embeddings have improved drastically, but mainstream IR evals have stagnated since MSMARCO + BEIR.
We ask: on private or tricky IR tasks, are rerankers better? Surely, reranking many docs is best?
It's time to revisit common assumptions in IR! Embeddings have improved drastically, but mainstream IR evals have stagnated since MSMARCO + BEIR.
We ask: on private or tricky IR tasks, are rerankers better? Surely, reranking many docs is best?
November 20, 2024 at 7:47 PM
Mat is not on 🦋—posting on his behalf!
It's time to revisit common assumptions in IR! Embeddings have improved drastically, but mainstream IR evals have stagnated since MSMARCO + BEIR.
We ask: on private or tricky IR tasks, are rerankers better? Surely, reranking many docs is best?
It's time to revisit common assumptions in IR! Embeddings have improved drastically, but mainstream IR evals have stagnated since MSMARCO + BEIR.
We ask: on private or tricky IR tasks, are rerankers better? Surely, reranking many docs is best?
Men's March Madness 2025: Reranking the remaining 8 teams 👀
Ranking Duke and the 7 teams standing in its way of a national title
We have our regional finalists. Let's rerank the eight remaining teams in the 2025 NCAA tournament.
www.espn.com
March 29, 2025 at 1:34 PM
Men's March Madness 2025: Reranking the remaining 8 teams 👀
listwise reranking with LLMs
November 18, 2024 at 7:47 PM
listwise reranking with LLMs
Reranking The World’s Billionaires By Wealth – And Altruism
Reranking The World’s Billionaires By Wealth – And Altruism
Forbes recalculated the wealth of the richest billionaires to add back in what they gave to charity. Here’s who moves up the ranks.
www.forbes.com
April 20, 2026 at 3:57 PM
Reranking The World’s Billionaires By Wealth – And Altruism
romance readers who have read Milla Vane's A HEART OF BLOOD AND ASHES might have a compelling reranking argument. just saying.
Mongers, ranked:
1. Cheese
2. Fish
3. Rumor
4. Iron
5. War
6. Conspiracy
1. Cheese
2. Fish
3. Rumor
4. Iron
5. War
6. Conspiracy
August 30, 2025 at 10:45 PM
romance readers who have read Milla Vane's A HEART OF BLOOD AND ASHES might have a compelling reranking argument. just saying.
LLM4Ranking
An Easy-to-use Framework of Utilizing Large Language Models for Document Reranking.
Repo: github.com/liuqi6777/ll...
An Easy-to-use Framework of Utilizing Large Language Models for Document Reranking.
Repo: github.com/liuqi6777/ll...
April 14, 2025 at 6:36 AM
LLM4Ranking
An Easy-to-use Framework of Utilizing Large Language Models for Document Reranking.
Repo: github.com/liuqi6777/ll...
An Easy-to-use Framework of Utilizing Large Language Models for Document Reranking.
Repo: github.com/liuqi6777/ll...
New in Science: Reranking partisan animosity in algorithmic social media feeds alters affective polarization www.science.org/doi/10.1126/...
Key from @jennyallen.bsky.social @jatucker.bsky.social: (1) important new paradigm (2) but not clear why results differ
www.science.org/doi/full/10....
Key from @jennyallen.bsky.social @jatucker.bsky.social: (1) important new paradigm (2) but not clear why results differ
www.science.org/doi/full/10....
December 1, 2025 at 3:38 PM
New in Science: Reranking partisan animosity in algorithmic social media feeds alters affective polarization www.science.org/doi/10.1126/...
Key from @jennyallen.bsky.social @jatucker.bsky.social: (1) important new paradigm (2) but not clear why results differ
www.science.org/doi/full/10....
Key from @jennyallen.bsky.social @jatucker.bsky.social: (1) important new paradigm (2) but not clear why results differ
www.science.org/doi/full/10....
🌐 I've just released Sentence Transformers v5.4: we're going fully multimodal for embeddings & reranking!
Also featuring a modular CrossEncoder, and automatic Flash Attention 2 input flattening.
Highlights in 🧵
Also featuring a modular CrossEncoder, and automatic Flash Attention 2 input flattening.
Highlights in 🧵
April 9, 2026 at 1:55 PM
🌐 I've just released Sentence Transformers v5.4: we're going fully multimodal for embeddings & reranking!
Also featuring a modular CrossEncoder, and automatic Flash Attention 2 input flattening.
Highlights in 🧵
Also featuring a modular CrossEncoder, and automatic Flash Attention 2 input flattening.
Highlights in 🧵
An extremely fast self-hosted alternative to Jev built on the #ModernBERT #embedding #model. Laya is multilingual and #typesafe, #python.
#OpenSource Apache 2.0 license.
#AI #LLM #Reranking
GitHub github.com/NandhaKishor...
#OpenSource Apache 2.0 license.
#AI #LLM #Reranking
GitHub github.com/NandhaKishor...
September 22, 2026 at 4:13 PM
An extremely fast self-hosted alternative to Jev built on the #ModernBERT #embedding #model. Laya is multilingual and #typesafe, #python.
#OpenSource Apache 2.0 license.
#AI #LLM #Reranking
GitHub github.com/NandhaKishor...
#OpenSource Apache 2.0 license.
#AI #LLM #Reranking
GitHub github.com/NandhaKishor...
Reranking: Third. Second.
The sexiest thing is bringing me on to facilitate reviewer relationships, generate value, and champion authors.
The sexiest thing is bringing me on to facilitate reviewer relationships, generate value, and champion authors.
The second sexiest thing a publisher can do is have a list of which publicists are working on which titles so I know who to talk too. That's because the sexiest thing that a publisher can do is hire enough people so Marketing & Publicity has multiple staff members.
January 29, 2025 at 9:11 PM
Reranking: Third. Second.
The sexiest thing is bringing me on to facilitate reviewer relationships, generate value, and champion authors.
The sexiest thing is bringing me on to facilitate reviewer relationships, generate value, and champion authors.
Whoa, excellent study just dropped in Science!
"Reranking partisan animosity in algorithmic social media feeds alters affective polarization"
www.science.org/doi/10.1126/...
Led by @tiziano.bsky.social and @msaveski.bsky.social
"Reranking partisan animosity in algorithmic social media feeds alters affective polarization"
www.science.org/doi/10.1126/...
Led by @tiziano.bsky.social and @msaveski.bsky.social
Reranking partisan animosity in algorithmic social media feeds alters affective polarization
Today, social media platforms hold the sole power to study the effects of feed-ranking algorithms. We developed a platform-independent method that reranks participants’ feeds in real time and used thi...
www.science.org
November 30, 2025 at 1:04 AM
Whoa, excellent study just dropped in Science!
"Reranking partisan animosity in algorithmic social media feeds alters affective polarization"
www.science.org/doi/10.1126/...
Led by @tiziano.bsky.social and @msaveski.bsky.social
"Reranking partisan animosity in algorithmic social media feeds alters affective polarization"
www.science.org/doi/10.1126/...
Led by @tiziano.bsky.social and @msaveski.bsky.social
We found a surprising result when measuring embeddings and rerankers on internal evals.
1- Reranking few docs improves recall (expected).
2- Reranking many docs degrades quality (!).
3- Reranking too many documents is quite often worse than using embedding model alone (!!).
1- Reranking few docs improves recall (expected).
2- Reranking many docs degrades quality (!).
3- Reranking too many documents is quite often worse than using embedding model alone (!!).
November 20, 2024 at 7:47 PM
We found a surprising result when measuring embeddings and rerankers on internal evals.
1- Reranking few docs improves recall (expected).
2- Reranking many docs degrades quality (!).
3- Reranking too many documents is quite often worse than using embedding model alone (!!).
1- Reranking few docs improves recall (expected).
2- Reranking many docs degrades quality (!).
3- Reranking too many documents is quite often worse than using embedding model alone (!!).
Reranking after Operation Italy
1. Rachel
2. Sam
3. Sue
4. Teeny
5. Genevieve
6. Andy
1. Rachel
2. Sam
3. Sue
4. Teeny
5. Genevieve
6. Andy
Potential survivor 47 winner ranking based on current standing (threat index, challenge results, voting accuracy, and a small personal bias bonus):
1. Rachel
2. Sam
3. Andy
4. Genevieve
5. Sue
6. Kyle
7. Caroline
8. Teeny
1. Rachel
2. Sam
3. Andy
4. Genevieve
5. Sue
6. Kyle
7. Caroline
8. Teeny
December 10, 2024 at 2:55 AM
Reranking after Operation Italy
1. Rachel
2. Sam
3. Sue
4. Teeny
5. Genevieve
6. Andy
1. Rachel
2. Sam
3. Sue
4. Teeny
5. Genevieve
6. Andy
This article provides more data showing how easy it is to manipulate humans on social media.
These researchers rerouted the algorithm on Twitter to push some users toward “antidemocratic attitudes and partisan animosity”.
It only took 1 week to elicit changes that used to take 3 years.
These researchers rerouted the algorithm on Twitter to push some users toward “antidemocratic attitudes and partisan animosity”.
It only took 1 week to elicit changes that used to take 3 years.
Reranking partisan animosity in algorithmic social media feeds alters affective polarization
Today, social media platforms hold the sole power to study the effects of feed-ranking algorithms. We developed a platform-independent method that reranks participants’ feeds in real time and used thi...
www.science.org
November 28, 2025 at 2:42 PM
This article provides more data showing how easy it is to manipulate humans on social media.
These researchers rerouted the algorithm on Twitter to push some users toward “antidemocratic attitudes and partisan animosity”.
It only took 1 week to elicit changes that used to take 3 years.
These researchers rerouted the algorithm on Twitter to push some users toward “antidemocratic attitudes and partisan animosity”.
It only took 1 week to elicit changes that used to take 3 years.
AI interview prep covers context windows for RAG via top-k retrieval and reranking, temperature settings for deterministic outputs, model weights, and training versus inference differences.
September 27, 2026 at 4:28 PM
AI interview prep covers context windows for RAG via top-k retrieval and reranking, temperature settings for deterministic outputs, model weights, and training versus inference differences.
Yep, platform algorithms feed polarization - monetizing rage, grievances, hate...
"Up-ranking increased political polarization, whereas down-ranking decreased it."
Study: Reranking partisan animosity in algorithms alters polarization www.science.org/doi/10.1126/...
"Up-ranking increased political polarization, whereas down-ranking decreased it."
Study: Reranking partisan animosity in algorithms alters polarization www.science.org/doi/10.1126/...
Reranking partisan animosity in algorithmic social media feeds alters affective polarization
Today, social media platforms hold the sole power to study the effects of feed-ranking algorithms. We developed a platform-independent method that reranks participants’ feeds in real time and used thi...
www.science.org
December 2, 2025 at 11:09 PM
Yep, platform algorithms feed polarization - monetizing rage, grievances, hate...
"Up-ranking increased political polarization, whereas down-ranking decreased it."
Study: Reranking partisan animosity in algorithms alters polarization www.science.org/doi/10.1126/...
"Up-ranking increased political polarization, whereas down-ranking decreased it."
Study: Reranking partisan animosity in algorithms alters polarization www.science.org/doi/10.1126/...
Können Algorithmen wirklich beeinflussen, wie polarisiert unsere Gesellschaft ist?
Eine neue Studie hat nun selbst gemessen, dass ein Algorithmus, der „antidemokratische Einstellungen und parteipolitische Feindseligkeit“ im Feed stark einblendet, die Polarisierung antreibt
Siehe:
Eine neue Studie hat nun selbst gemessen, dass ein Algorithmus, der „antidemokratische Einstellungen und parteipolitische Feindseligkeit“ im Feed stark einblendet, die Polarisierung antreibt
Siehe:
Reranking partisan animosity in algorithmic social media feeds alters affective polarization
Today, social media platforms hold the sole power to study the effects of feed-ranking algorithms. We developed a platform-independent method that reranks participants’ feeds in real time and used thi...
www.science.org
December 3, 2025 at 8:44 AM
Können Algorithmen wirklich beeinflussen, wie polarisiert unsere Gesellschaft ist?
Eine neue Studie hat nun selbst gemessen, dass ein Algorithmus, der „antidemokratische Einstellungen und parteipolitische Feindseligkeit“ im Feed stark einblendet, die Polarisierung antreibt
Siehe:
Eine neue Studie hat nun selbst gemessen, dass ein Algorithmus, der „antidemokratische Einstellungen und parteipolitische Feindseligkeit“ im Feed stark einblendet, die Polarisierung antreibt
Siehe:
Compared against CLEF-TAR 8 DTA used in Akinseloyin 2024/2026. It is slightly worse at WSS95 & better at WSS100 using essemble of Claude 3 + Gemini reranking. 2026 LLM esemble = GPT-4o Mini + Gemini 1.5 Flash + Claude 3 Haiku, with embedding reranking - not bad at all /1
September 24, 2026 at 4:42 PM
Compared against CLEF-TAR 8 DTA used in Akinseloyin 2024/2026. It is slightly worse at WSS95 & better at WSS100 using essemble of Claude 3 + Gemini reranking. 2026 LLM esemble = GPT-4o Mini + Gemini 1.5 Flash + Claude 3 Haiku, with embedding reranking - not bad at all /1
My interpretation of this article:
You don’t need a violent coup or a hacked voting machines to overthrow democracy.
You just need control of the algorithm.
You don’t need a violent coup or a hacked voting machines to overthrow democracy.
You just need control of the algorithm.
Reranking partisan animosity in algorithmic social media feeds alters affective polarization
Today, social media platforms hold the sole power to study the effects of feed-ranking algorithms. We developed a platform-independent method that reranks participants’ feeds in real time and used thi...
www.science.org
November 28, 2025 at 3:09 PM
My interpretation of this article:
You don’t need a violent coup or a hacked voting machines to overthrow democracy.
You just need control of the algorithm.
You don’t need a violent coup or a hacked voting machines to overthrow democracy.
You just need control of the algorithm.