Uhh I’m not Anisha but I saw something weird just now:
The 30 days don’t make any sense because this is the second time I’ve seen this card
No clue what causes this tbh.
1 Like
Agreed.
Definitely agree with this one and would classify as a bug - there is no reason for the Hard and Solid intervals to be the same (or any of the other responses).
I haven’t really had the time lately to actively test the Anki SM-2 Scheduler for the more sneaky bugs, so it could be why I don’t seem to have any pressing issues with it as of now. The most recent SRS-related bug/unimplemented feature (which should be a default feature imo) that I have come across is this one: Overdue Cards: Have Long Easy Intervals & Related · Issue #933 · remnoteio/remnote-issues · GitHub.
Unrelated:
Older issues include Next repetition time not showing any date/time, or showing a date/time from the past (usually indicates that the card was deactivated by itself somehow) - if I’m not wrong, this issue has been fixed for the most part, but there could be a subset of users that still come across this behavior.
Not 100% sure, but it looks to be the case in 88DM88’s screenshot:
1 Like
Nvm, the scheduler is right: Going by the tutorial the next interval is computed by multiplier*(last scheduled interval+days late/hardness divider)
In my example that would be (2.5 * 1.3) * (11/(60 * 24)+9d) which results in 29.270d so almost 30 days. The missing hours are probably because of the 9 days not being exact
1 Like
I don’t think so.
In reality, the metadata shows that your card was still in the Learning phase. So, that formula does not apply yet (it will apply when the card graduates to the Exponential phase.
In the learning phase, as the tutorial says:
All intervals in the learning phase are taken from a fixed preset that is independent of your reported scores (ex. 1 day, 2 days, then 4 days). You can graduate a card from this phase into the Exponential Phase by pressing “Easy”, or by finishing the fixed preset intervals.
I imagine that you use 11 min as your first step; I also would hardly believe that your second step is 30 days (it would be cool if you show us your settings). In thesis, overdueness, i.e. days late, should not affect you card when in learning; you would have no benefit from rating you card late while you are still in learning phase (while when in the exponential phase there is a benefit, because the algorithm understands that if you were capable to remember even after a higher interval than the one calculated, this knowledge is stronger in your brain). So, if the system is calculating days late when your card is still in the learning phase, it is behaving wrongly.
Supposing now that it was in the exponential phase (just for argumentation, as we know it was still in the learning phase), by the formula you presented, I imagine you use an interval modifier of 130%. Is it true? (it would be nice if you show us this setting also).
As you said,
Next interval = multiplier * ( Last scheduled interval + (days late/hardness divider) )
Including the effect of the Interval Modifier:
Next interval = [multiplier * ( Last scheduled interval + (days late/hardness divider) ) ] * 130%
Your multiplier is 2.5 (“Ease” shown in the metadata). Your last scheduled interval is 11 min (11/(60 * 24), as you showed. But, if you are pressing Good, your Hardness Divider would be 2, and the 9 days late would be halved. The tutorial again:
The Hardness divider is a fixed divider computed from your last score.
A last score of Hard has a divider of 4
A last score of Good has a divider of 2
A last score of Easy has a divider of 1
The formula would be (supposing you use an interval modifier of 130%):
Next interval =(2,5*(11/(60*24)+9/2))*1,3
This would result in 14,6 days.
So, for now (until we see your settings), all we can say in that the system should have used your second step as next interval (as the card was in the learning phase); and that even if is was in the exponential phase, and even if you changed the interval modifier from 100% to 130% (so increasing the spacing effect), the interval would still be half of that given by the system.
It still looks very wrong to me!
If @Martin could peak at this I would be grateful!
2 Likes
Yeah I forgot about the learning phase, sorry
but at least we know what calculation the scheduler probably did. Seems like there are multiple bugs at play that mess up the next interval by a lot
Ran into a similar issue like before, but this time my ease factor was influenced as well even though the card is in the learning phase. Maybe RemNote somehow mixed up the default scheduler with the anki scheduler because if I remember correctly the default scheduler uses a default ease of 2.3
Edit: another example
Once again wrong ease during the learing phase and only the easy button being affected. To see what would happen I clicked easy and this is the metadata of the card now:
2 Likes
Linking this bug report here:
1 Like
I recently reviewed some cards on my phone but the results haven’t really synced to the desktop. I checked this card and it doesn’t show any of the results that I did recently.
The scheduling is messed up again for me after trying to use a second device (web app only). I thought the main cause was the Anki SM-2 algorithm, but now it seems that the Global Default scheduler is unreliable as well. Unfortunately, this basically prevents me from using RemNote in its current state.
The picture below is an example of what is happening — this specific card was supposed to have an interval of 2 months but instead it gets presented after only 2 days.
All this seems very concerning to me, is it possible to have an “official” statement on those issues with the Queue please?
1 Like
I can confirm that I have issues with the “normal algorithm” as well. I have not used the Anki algorithm so far, but during the last months I started to realize some problems. Cards show up multiple times a day (although they should only reappear within 1-3 months) or intervals seem to be messed up. This is really frustrating, especially since we pay for this product.
I use RemNote very day for multiple hours, my queue is around 4-500 card a day and I practice around 200 cards a day.
I also realized that the count is not resetted the next day. Today for example RemNote just told me I have studies 399 cards, while in reality I only have studies 50 cards so far. Also the “days in a row” are not counted correctly.
At this point I am quite frustrated about it since I do rely on the product and I expected it to be reliable since it is now a paid product and not a beta anymore. However, quite often it still feels like a beta.
1 Like
@Martin has answered my issue posted in Too long scheduling intervals (Anki SM-2 shceduler) · Issue #1062 · remnoteio/remnote-issues · GitHub and in the beta server a new version is already running in which the too long intervals have been corrected; I’ve been testing it and so far could not detect misbehavior. But I have seen before this strange behavior of cards not yet due being shown in the queue. I expect this has also been solved in the last change that is still only in beta server but will be updated in the stable next version.
The issue you linked seems to only deal with the SM-2 algorithm bug, I’ve already switched completely away from it to the Global Default scheduler due to these problems. It doesn’t touch on the issues me and other users reported above. What makes you think it’s going to be solved in the next update? I am really worried that a lot of my cards have not been scheduled properly and it seems updates often fix some older issues with the scheduler but also introduce new ones.
I’re right. I just hope it may have been fixed, as for fixing the Anki SM-2 problems they had to visit the scheduling code, and it is possible that other issues have also been fixed. This is all that I meant. Further, I have tested the beta server version for a while, and I didn’t find this misbehavior anymore so far. This fact strengthens that hope I mentioned. I’ll keep looking, and if I find again the same problem you did (as I told I found in the past, in a few cards) I’ll post here, to help the team find the issue.
1 Like
Hello, it’s my first time actually writing in the forum, and I’m a pretty new user (maybe a month?) so I’m still learning how to use RemNote properly, I hope I can still contribute to the conversation.
I wanted to add my experience regarding this issue. I have the basic free version, so I only have the “normal” scheduler, but am experiencing similar problems as the other users above have mentioned:
- the same card shows up multiple times a day and several days in a row without respecting the intervals (for example a card I recalled easily and selected the smiley 7 days button for shows up in a few minutes or the next day)
- cards from folders that I suspended keep showing up in my queue, also without respecting intervals
1 Like
thank you all for the contributions, if someone is interested, can you go through this conversation and come up with an outline of the different issues that are being faced here and connecting it to the specific posts for that? That would just be super helpful for us in understanding and I can ask Martin to jump head on in tackling those issues then.
@moderators would appreciate your help too, if anyone is a heavy queue user and understands these issues.
Hello Karthikk, hello all
It’s been a while I’m lurking here and on discord and I’m using remnote since several months, not a pro/lifelong learner because I consider the product is not reliable and finished yet but this is not the topic here.
Basically Remnote is a note-taking app with a SRS built inside.
I write this because I’m very concerned, reading all this discussion it seems the Queue is broken, or worse than this, maybe never fully fonctionned.
The problem is known since january at least, more than 3 months ago. There are several testimony and examples but yet still no fix or official announcement about it.
Karthikk, not sure what you are asking here, the contributions seems crystal-clear to me and you already told them about Martin 3 weeks ago.
So, what now? I think they, we deserve transparency here, a core function of remnote is buggy and nothing is done, at least this is the official (no) communication coming from your team.
I repeat, from an “outsider” this is very alarming. Please, I urge you to act about it, thank you.
3 Likes
I should have probably foreseen this, and I am to blame for causing concern. Maybe if I explained more explicitly what made me begin this thread it could help alleviate some concerns?
I started this thread not because the queue was broken, it was because I had been noticing that certain users seemed to be noticing some odd behaviours in the queue and it was hard to reproduce these issues, as they were probably edge cases and sometimes account specific even.
This thread was meant to be a lightning rod for such concerns, where a user could come and share these odd findings that they weren’t able to pinpoint or catch (as they would make their way to the GitHub tracker and get fixed), in hopes that someone else also faced similar experiences but were unsure if they should report it.
However, I agree that SRS related issues need to be dealt with more priority than other matters. The issue has been that being a small team (Martin still is technically the only one who can work on bugs, others are still in training/hiring phase), we had trouble prioritising. I have been cleaning the bug reports on forum and GitHub only because it could give us a more clear idea about what’s hurting us the most.
TL;DR - We have almost identified and gathered all queue related issues, Martin is currently going through them. The main focus now and for the near future is fixing these bugs and improving design and UX. A handful of queue related bugs have already been fixed in the beta server.
You will be able to see what has been done, once the next update is released (through the changelog).
The majority of the problems are around
- Queue metadata Syncing between devices
- Some scheduler abnormalities (we have had some success with the bugs related to this category)
- UI related queue bugs (low priority for now)
We hope to find them all soon, and this thread has been very instrumental in understanding these issues, so thank you everyone for that!
3 Likes