At the bottom of the list is a summary of the number of "best papers" by institution. Interestingly, Microsoft Research is at the top of the list:

    Microsoft Research                     32.4
    Stanford University                    26.8
    Massachusetts Institute of Technology  24.6
    University of Washington               24.1
    Carnegie Mellon University             22.9
    University of California Berkeley      19.5
    ...

MSR attracts some of the best researchers and makes it easy to get research done. No teaching requirements, grant applications or scrambling for funding. Just research. I sincerely hope tomorrow's Microsoft keeps MSR well-funded, it's a benefit to the whole community.

MSR is very well-regarded, rather autonomous from Microsoft's product division, and likely larger than any university's CS department. It's not much easier to get a principal researcher there than a professorship at a great university.

They're not as broad as most universities, but they pretty reliably have the most papers in top systems conferences like OSDI and SOSP.

I understand what are you trying to say. But this is a not a reasonable comparison.

And by a nice margin. IBM is quite far below, I'd never expect that.

IBM would gain a bit if computer architecture conferences were included.

True. Even though, first I got confused between papers and patents (they have a tradition of submitting lots of them) and last I wonder if they didn't slow low level research, if not then they keep it quiet, unlike a few years back where they'd talk about they latest PowerPC feats.

Just a small caveat: There is some evidence that best paper awards do not really correlate with great influence of those papers, as measured by the number of citations a paper receives afterwards [1].

[1] http://www.bartneck.de/publications/2009/scientometricAnalys...

Too bad SIGGRAPH doesn't give such an award; I was curious to look at the top graphics papers.

Not top papers, but you can read many of the accepted papers by following the author links here: http://kesen.realtimerendering.com/sig2013.html

Thanks! I was interested in seeing the list of "best papers" just to see what the most interesting or innovative research over the years has been, but this is all good stuff too.

Looking at Google scholar results might be interesting:

http://scholar.google.com/citations?hl=en&vq=eng_computergra...

I know this topic has been done to death but it still blows my mind that most of those papers are not freely available. It would be nice if the freely available ones were marked as such in the list.

It still blows my mind that people claim this is true. Almost all CS papers are on the author's website, eprint, arxiv, etc. Google scholar even links to them.

Don't get be wrong, closed publications are a travesty. But they are a travesty because institutions pay out of government grant money (indirectly via overhead) to pay for subscriptions and because it's government funded research. They are not, at least in anything approaching the average case, a travesty because researchers cannot access results.

Really? I just clicked 10 random link and all Google Scholar searches redirected to a http://dl.acm.org/ paywall. Maybe I'm missing something?

edit: Ah, I just noticed some searches have a direct link to the PDF next to the main search result "[PDF] from domain".

The primary link is almost always to the canonical version which is usually the pay walled one(even if it's not the most in depth version the author wrote). Did you check for the links on the right hand side? Those are the pdf's that google found. Of a pseudo random set I clicked on, all had links.

If you search little bit more, you would be be able to find that work somewhere else for free download. Some application paper published by Springer etc are difficult to get.

If you are a member of a public library in a metropolitan area, take a look at the available journals and journal databases there. You might be pleasantly surprised.

It's not "freely" available but it's not something that everyone has to pay publishers directly for.

Hmm. Most that I clicked on were freely available. The links go to Google Scholar searches, and most of those indicate that a PDF is available.

For a lot of these links, I am not able to find the pdf as they are behind signups or paywall. sad state of affairs

Interestingly, "The Anatomy of a Search Engine" was not on this list.

http://infolab.stanford.edu/~backrub/google.html

An interesting list. I'd enjoy seeing this in other fields besides CS.

I'm not sure how far afield from CS you actually want to go, but the list for music theory is at http://societymusictheory.org/archive/publications.

I would be interested in physical/chemical sciences MRS, APS and ACS come to mind as examples of things with annual meetings which might have best paper awards.

As a biologist, where we emphasize publication over conferences, I am really curious as to what "best" means? Innovative? Most data to back hypothesis?

In CS, conference papers act much like journal papers (though often quite a bit shorter). Unlike journal papers, a talk or poster (by one of the authors) is also required to be presented at the conference in order to have the paper published. Though there are exceptions, the majority of these conferences are published in either ACM or IEEE libraries.

The Best Paper Award is typically some combination of a decision by the Program Committee, aggregated votes by attendees of the conference, and input from the paper reviewers who originally peer-reviewed the work.

There isn't a single metric for the award, and it can be quite subjective: sometimes it is an innovate topic, sometimes it is simply an interesting perspective, sometimes it is a well-executed presentation (even if the research itself wasn't all that interesting), sometimes it is because the authors used a difficult-to-acquire or real-world data set (internal Google or Microsoft data, game data directly from Blizzard), sometimes it is because of a challenging participant base (for example, interviewing minorities in elementary schools or a longitudinal study tracking participants over a decade), and sometimes it is simply excellent experimental design and statistical analysis.

In support of barik's comment, I want to add that conference papers in computer science are peer reviewed, and the top conferences have an acceptance rate that is usually less than 15%. Simply, conference papers, rather than journal articles, is where most innovative computer science work goes. This is very true for systems, languages, software engineering, databases and HCI, but less so for the more theoretical areas.

This topic comes up when talking with researchers outside of computer science so often that the Computer Science Research Association wrote up a memo explaining it: http://cra.org/resources/bp-view/evaluating_computer_scienti...

Peer review doesn't necessarily mean anything objective unless it is totally doubly-blinded, which, ironically, is not used by many top system conferences.

What's the best tablet upon which to actually read all these papers?

An excellent list! Love it. Did you compile it manually?

Reposting a worthy comment from user chimmy, who has been hellbanned for 1.5 years for no apparent reason:

> This is a great list but I would rather look at the most cited papers from that conference (say 10 years later). As an example, MapReduce did not win the best paper in OSDI 2004. However, it has impacted the industry like no other paper in that conference.

You can get some of this from Google Scholar's metrics pages, which list, for each venue, the most-cited papers since 2008. These indeed seem to be different than the best-paper selections, though I haven't examined this systematically. (I guess you could take the best-paper and look at its citation rank out of papers that year. But this raises other questions; for example, is the point of best paper award to select the paper that will be most influential?)

From this page, http://scholar.google.com/citations?view_op=top_venues&hl=en... , click "Subcategories" under "Engineering & Computer Science": for example, "Artificial Intelligence," "Computational Linguistics," or "Human Computer Interaction." It lists the most-cited-venues for each area, but if you click on a particular venue you get its list of most-cited-papers.

Here are a few that are also on Jeff Huang's list:

ICML: http://scholar.google.com/citations?hl=en&vq=eng_artificiali...

NIPS: http://scholar.google.com/citations?hl=en&vq=eng_artificiali...

AAAI: http://scholar.google.com/citations?hl=en&vq=eng_artificiali...

ACL: http://scholar.google.com/citations?hl=en&vq=eng_computation...

EMNLP: http://scholar.google.com/citations?hl=en&vq=eng_computation...

CHI: http://scholar.google.com/citations?hl=en&vq=eng_humancomput...

CVPR: http://scholar.google.com/citations?hl=en&vq=eng_computervis...

A site that lists both (building off this one) is http://arnetminer.org/conferencebestpapers

What impacts industry 10 years later is more a measure of marketing and brand association than academic excellence.

Google could fart now and the internet will still cry roses.

The same research/algorithms/software produced by an independent individual or lab would simply die in obscurity.

That's because, unlike physical science and engineering disciplines, the application of discoveries in computer science (i.e. software engineering) is done in the context of an ultra-competitive popularity-contest culture, mixed with a strong strain of loner-nerd-hacking-the-gibson worship.

I came to software development from the physical sciences (university level academic and theoretical as well as applied/engineering). To me the differences were abundant and, frankly, shocking. For a discipline that fancies itself one of the more intellectual, there seems to be an awful lot of petty one-upsmanship and "john galt genius" idiocy at work.

Oh crying baby, don't cry.

Hmm... TL;DR?

Huh? It's a list of best papers. What kind of summary can you give of hundreds of CS papers in different fields? "There sure is a lot of stuff you can do with computers!"

greenyoda already included the first few lines of the summary of institutions which contributed most to this list, which is about as close to a TL;DR as you can get for this link: https://news.ycombinator.com/item?id=6992714

But really, what were you expecting a TL;DR to be? There's no way you can summarize nearly two decades worth of computer science research in a single TL;DR.

cough Irony cough

Wooowee, did I ever strike a nerve :)